Species Identification and Habitat Assessment of the South Texas Siren

Home , Greater siren, Hydrophilus triangularis, Lesser siren, Palaemonetes paludosus, Sirenidae, Siren (genus)

University of Texas Rio Grande Valley ScholarWorks @ UTRGV

UTB/UTPA Electronic Theses and Dissertations Legacy Institution Collections

7-2015

Species Identification and Habitat Assessment of the South Texas Siren

Taylor C. LaFortune The University of Texas Rio Grande Valley

Follow this and additional works at: https://scholarworks.utrgv.edu/leg_etd

Part of the Animal Sciences Commons, and the Biology Commons

Recommended Citation LaFortune, Taylor C., "Species Identification and Habitat Assessment of the South exasT Siren" (2015). UTB/UTPA Electronic Theses and Dissertations. 4. https://scholarworks.utrgv.edu/leg_etd/4

This Thesis is brought to you for free and open access by the Legacy Institution Collections at ScholarWorks @ UTRGV. It has been accepted for inclusion in UTB/UTPA Electronic Theses and Dissertations by an authorized administrator of ScholarWorks @ UTRGV. For more information, please contact [email protected], [email protected].

Species identification and habitat assessment of the South Texas siren

A Thesis Presented to the

Faculty of the College of Science, Mathematics, and Technology

The University of Texas at Brownsville

In Partial Fulfillment

Of the Requirements for the Degree

Master of Science

Taylor Courtney LaFortune

July 2015

Taylor Courtney LaFortune

July 2015

ACKNOWLEDGEMENTS

Thank you to the Texas Parks and Wildlife Department for funding this thesis research. First and foremost, I would not have been able to complete this project without the knowledge, guidance, and support of my advisor Dr. Richard J. Kline. I would also like to thank my committee member Dr. Daniele Provenzano at The University of Texas at Brownsville for his encouragement, advice, and humor throughout the entire process, and Dr. Andrew Gluesenkamp at Texas Parks and Wildlife for his insightfulness and direction. A special thanks also goes to Dr. Gluesenkamp, Dan Saenz, Travis LaDuc,

Heidi Smith-Parker and Paul Moler for providing tissue samples. I appreciate the help of all the volunteers, especially Kristen Kline, Maria Cooksey, and Crystal Lopez, for their constant willingness and enthusiasm for assisting with field work during the summer heat in Texas, and Connie Mata, who spent long days and nights working in the laboratory.

Without the support of my friends and volunteers, this research would not have been possible. I would also like to thank the numerous land owners, state park officials, and the US Fish and Wildlife agency for assisting me with the permits and study sites for conducting my fieldwork. Lastly, I would whole-heartedly like to thank my parents, Marc and Ann LaFortune, my sisters, Andrea LaFortune and Marquis Stakes, and Adam

Duffek, for their constant love, advice, and unwavering support.

iii

LaFortune, T.C., “Species identification and habitat assessment of the South Texas siren.” Unpublished Master of Biology Thesis, The University of Texas at Brownsville, Brownsville, TX, 2015.

ABSTRACT

At least one species of siren is endemic to the Rio Grande Valley of Texas, yet very little is known about the populations of the region. Texas Parks and Wildlife refers to the populations of Siren spp. in South Texas as the “South Texas siren (large form Siren sp. 1),” and recognizes these populations as a threatened species, yet their current population status remains unclear. The species identification of the South Texas siren has been hampered by similarity in morphology across species, and by the lack of complete siren genetic sequences in the NCBI GenBank database. In addition to species ambiguity, very little is known about the preferred habitat characteristics of sirens, specifically in South Texas. The aim of this study was to identify the species of Siren spp. that inhabit South Texas, and to assess the vegetation and environmental variables of siren habitat. Sirens were collected from seventeen water bodies throughout South Texas. Thirty-six sites were assessed for siren presence and correlation with environmental variables, co- occurring species, and vegetation composition. There was no significant correlation between siren presence and the environmental factors; however, nearly all sirens were collected in water bodies that had a high (>95%) percent cover of edge vegetation, and siren abundance appeared to be affected by seasonality. A total of 65 South Texas siren tissue samples were collected between 2013 and 2015. Confirmed specimens of Siren lacertina were compared with the South Texas siren samples, to analyze both coding and non-coding regions (protein coding genes, rRNAs, and tRNAs). For species identification, nine complete mitochondrial genomes were sequenced, and comparisons were made against single genes to assess their utility for species resolution. Sequence divergence and phylogenetic relationships suggest that siren populations in South Texas are composed of at least one distinct species that differs from published sequences for Siren intermedia and S. lacertina. In addition, the results suggest that CO1 is likely the most useful gene for species identification in lieu of the complete mitochondrial genome. The results from this study will provide critical information for this cryptic species, and will aid in the development of future conservation and management practices.

Table of Contents ACKNOWLEDGEMENTS ...... III ABSTRACT ...... IV TABLE OF CONTENTS ...... V LIST OF TABLES ...... VII LIST OF FIGURES ...... VIII LIST OF ABBREVIATIONS ...... XI I. INTRODUCTION ...... 1

AMPHIBIAN DECLINE ...... 1 GENUS SIREN ...... 3 Habitat ...... 3 Seasonality ...... 6 Distribution ...... 8 SPECIES IDENTIFICATION ...... 9 Mitochondrial DNA ...... 10 OBJECTIVES AND HYPOTHESES ...... 14 II. MATERIALS AND METHODS ...... 16

STUDY SITES ...... 16 TRAP COLLECTIONS AND ENVIRONMENTAL VARIABLES ...... 16 PHOTO-ID ...... 19 SAMPLE COLLECTION ...... 20 HABITAT ANALYSIS ...... 20 Co-occurring Species Composition ...... 20 Vegetation Composition ...... 21 MTDNA ANALYSIS ...... 22 DNA extraction, cloning, and sequencing ...... 22 Sequence Assembly ...... 24 Sequence Alignments ...... 25 PHYLOGENETIC ANALYSIS ...... 26 Bayesian Analyses ...... 26 Maximum Likelihood Analyses ...... 28 Sequence Divergence Analyses ...... 28 III. RESULTS ...... 31

SIREN COLLECTION DATA ...... 31 WILD-ID ...... 32 HABITAT DATA ...... 33 Co-occurring Species Composition ...... 33 Environmental Variables ...... 34 Vegetation Composition ...... 34 PHYLOGENETIC DATA ...... 35 mtDNA content ...... 35 CMGS phylogeny and sequence divergence ...... 36 Gene phylogeny and sequence divergence ...... 39 Isolation by Distance ...... 42 IV. DISCUSSION ...... 44

SPECIES IDENTIFICATION ...... 44 DISTRIBUTION ...... 47 HABITAT CHARACTERISTICS ...... 48 V. CONCLUSION...... 52 VI. LITERATURE CITED ...... 53

TABLES…………………………………………………………………………………………………61 FIGURES……………………………………………………………………………………………..…64 SUPPLEMENTARY TABLES AND FIGURES……………………………………...……...86

LIST OF TABLES

Table 1. List of mtDNA primers used to amplify three large fragments of the siren mitochondrial genome for cloning, and primers used to primer walk within those three amplicons. Alternate primers are denoted by an asterisk (*). Bold primer names indicate the primers used for fragment amplification.

Table 3. Nucleotide composition of complete Siren mtDNA sequences from this study.

vii

LIST OF FIGURES

Figure 1. The state of Texas with all sampling locations (yellow stars) utilized for habitat assessment and environmental analysis. The circles outlining sites represent the nine regions defined based on geographic distance. White regions contain multiple sampling sites, and red regions contain only one sampling site (excluded from habitat analysis). Dashed circles represent regions with siren presence. All locations were within (C) Cameron, (H) Hidalgo and (K) Kleberg Counties.

Figure 2. The state of Texas with all locations of siren collections from Texas utilized for mtDNA analysis in this study. Figure 3. Locations of primers for primer walking within the three fragments cloned for sequencing the Siren spp. complete mitochondrial genome. Arrows represent the 5’3’ direction of the nucleotide sequence. Primer sequences are displayed in Table 1.

Figure 4. Gene organization and cloning fragment locations for the complete mitochondrial genome of Siren spp. Forward arrows represent genes encoded on the heavy strand, and reverse arrows represent genes encoded on the light strand. Black arrows represent protein-coding genes, white arrows represent ribosomal subunits, and the line displays the location of the intergenic spacer (IGS) and non-coding region (D- Loop).

Figure 5. Histogram of the total siren abundance collected (regardless of site location) during each month of collection in this study from 2013 to 2015. Abundances display siren collection from systematic sampling, opportunistic captures, and the CPUE study. Figure 6. Histogram of the catch-per-unit-effort (CPUE) of sirens collected during each month of sampling (regardless of site) in this study from 2013 to 2015. CPUE values display only sirens collected from systematic trapping (not opportunistic captures). Figure 7. Variation in body coloration and spot pattern of three sirens collected in South Texas in this study. Figure 8. Plot of the weight-length relationships and log-regression for all sirens collected in South Texas from this study. A logarithmic trend-line produced a R2 = 0.93 correlation. Figure 9. Siren tail shapes of known Siren lacertina (A, B) and tails shapes of sirens collected in South Texas in this study (C –F). Figure 10. A small (18.5 g, 185 mm) siren collected from South Texas in October 2014 from this study, exhibiting a distinct yellow pattern along the side of the body.

viii

Figure 11. Variation in siren head spot pattern and coloration of known Siren lacertina (A) and sirens collected in South Texas in this study (B-D). Figure 12. Wild-ID photograph recognition software analysis for the comparison of the spot patterns on the dorsal region of the head among all sirens within the database. The red box displays the comparison value of 0.0245, which is not considered a match (< 0.100). Figure 13. Non-metric multidimensional scaling (MDS) ordination plot of the combined percent cover (from the Aquatic, Emergent and Edge zones) for thirty-six sites from May to October 2014 (no samples in August or September) based on the presence/absence of sirens. No significance was observed between siren presence and percent cover. Low stress of 0.08 indicates that the 2D representation of the MDS plot was appropriate. The closer two points are to one another, the more similar the percent cover of all zones is between the sites. Figure 14. Non-metric multidimensional scaling (MDS) ordination plot of the percent cover in the Edge zone for thirty-six sites from May to October 2014 (no samples in August or September) based on the presence/absence of sirens Low stress of 0.01 indicates that the 2D representation of the MDS plot was appropriate. The closer two points are to one another, the more similar the Edge percent cover is between the sites. The percent edge cover was not significant, but almost all sites with siren collection had ≥ 95% cover.

Figure 15. Bayesian and Maximum Likelihood consensus tree of the siren+outgroup dataset (20 tRNAs, 2 rRNAs, 13 proteins). Support values above branches are the Maximum Likelihood (ML) bootstrap values (values < 50% not shown) and below the branches are the Bayesian Posterior Probabilities (PP), indicating support for the node. Asterisks (*) denote a node with support values of 100 for the ML support and 1.0 for PP support. Figure 16. Bayesian and Maximum Likelihood consensus tree of the siren-only complete mitochondrial genome dataset (22tRNAs, 2 rRNAs, 13 proteins, intergenic spacer, D- Loop). Support values above branches are the Maximum Likelihood bootstrap values (values < 50% not shown) and below the branches are the Bayesian Posterior Probabilities, indicating support for the node. Figure 17. Bayesian and Maximum Likelihood consensus topology tree of the phylogenetic relationships for all Siren spp. from Texas for the 16S dataset. Support values above branches are the Maximum Likelihood bootstrap values (values < 50% not shown) and below the branches are the Bayesian Posterior Probabilities, indicating support for the node. Figure 18. Bayesian and Maximum Likelihood consensus topology tree of the phylogenetic relationships for all Siren spp. from Texas for the CO1 dataset. Support values above branches are the Maximum Likelihood bootstrap values (values < 50% not

shown) and below the branches are the Bayesian Posterior Probabilities, indicating support for the node. Figure 19. Bayesian and Maximum Likelihood consensus topology tree of the phylogenetic relationships for all Siren spp. from Texas for the Cyt-b dataset. Support values above branches are the Maximum Likelihood bootstrap values (values < 50% not shown) and below the branches are the Bayesian Posterior Probabilities, indicating support for the node. Figure 20. Bayesian and Maximum Likelihood consensus topology tree of the phylogenetic relationships for all Siren spp. from Texas for the intergenic spacer dataset. Support values above branches are the Maximum Likelihood bootstrap values (values < 50% not shown) and below the branches are the Bayesian Posterior Probabilities, indicating support for the node. Figure 21. The correlation between pairwise comparisons of genetic distance (uncorrected p-values) and geographic distance (km) for the CO1 dataset for all Texas sirens (n = 45) collected in this study. The Mantel Test showed significant correlation with a p-value of 0.02 (p < 0.05). The linear correlation is r2 = 0.28. Figure 22. The collection locations in Texas for the eight Texas sirens sequenced for the complete mitochondrial genome. Yellow stars represent the common South Texas genotype, and the colored stars represent the three divergent sirens. The blue star represents the collection location of siren ATT-1, the green star siren SS23, and the red star siren SS20. Due to proximity of location, the eighth star (yellow) is not visible, but is located with the green star.

LIST OF ABBREVIATIONS

mtDNA Mitochondrial DNA

CMG Complete Mitochondrial Genome

CO1 Cytochrome oxidase 1

Cyt b Cytochrome b

IGS Intergenic Spacer bp Base Pairs

ML Maximum Likelihood

BS Bootstrap

PP Posterior Probabilities

AIC Akaike information criterion

GTR General Time Reversal

CPUE Catch Per Unit Effort

IBD Isolation by Distance

I. INTRODUCTION

Amphibian decline Amphibians have been undergoing drastic global decline throughout the 21st century (Pechmann and Wilbur, 1994; Stuart, 2004). In fact, an estimated 41% of amphibian species are considered threatened (The IUCN Red List of Threatened Species,

2015). Though amphibian communities experience natural population fluctuations, their sensitivity to environmental changes makes them particularly prone to population decline as a result of urbanization, habitat fragmentation, and other anthropogenic factors (James

P. Collins and Andrew Storfer, 2003; Stuart, 2004; Tipton et al., 2012). Despite the amphibian decline, a vast increase in the discovery of new amphibian species has been reported in the last few decades (Fouquet et al., 2007; Funk et al., 2012; Köhler et al.,

2005). This increase is primarily a consequence of increased exploration into tropical and high diversity regions, a rise in effective molecular tools, and the shifting characterization of the species concept (Fouquet et al., 2007; Grosjean et al., 2015). Anurans, an amphibian Order that includes frogs and toads, in neotropical settings, have primarily contributed to the new and cryptic species discoveries; however, many amphibians, such as salamanders, continue to be overlooked (Fouquet et al., 2007; Funk et al., 2012).

Despite comprising 8.5% of amphibian species today, members of the Order

Caudata remain an understudied group (Petranka, 2010). In the face of population decline, this lack of robust data and the general understanding of salamanders could be detrimental for viable conservation efforts (Funk et al., 2012; Grosjean et al., 2015).

More research is needed to understand the habitat and distribution of salamander species

and to delimitate cryptic species, so effective conservation and management practices can be implemented (Fouquet et al., 2007; Funk et al., 2012; Grosjean et al., 2015). This is especially true for the more cryptic salamander groups, including the genus Siren.

Sirens (Siren spp.) are nocturnal aquatic salamanders with a long, slender eel-like body, small forelimbs with four toes, and the paedomorphic characteristic of branched external gills (Petranka, 2010; Tipton et al., 2012). Two species of Siren, the greater siren

(Siren lacertina) and lesser siren (Siren intermedia), inhabit overlapping distributions throughout the Southeastern United States. Historically, both species have been documented in the Southernmost region of Texas; however, due to nearly indistinguishable morphology, much debate resides around the accuracy of siren species identification in South Texas (Oscar Flores Villela and Ronald A. Brandon, 1992).

Herpetologists throughout Texas offer speculative suggestions regarding species identification, but to date, little research has been conducted to resolve the ongoing debate. As a result, the populations of siren in South Texas have been termed the South

Texas siren (Large form sp. 1) by Texas Parks and Wildlife Department, and are considered a threatened species (NatureServe, 2013), despite being taxonomically unidentified.

Not only is the South Texas siren “species” perplexing in itself, but the scientific data for habitat preference, seasonality, and distribution is minimal. With the rise in resaca (ox-bow lake) restoration efforts throughout South Texas, and potential habitat alteration, the sirens in this region require immediate population assessment. In the following chapters, I present the scant scientific knowledge available for sirens, and address the necessary objectives for closing the knowledge gaps surrounding this genus,

specifically for South Texas. Primarily, I will address potential environmental variables that may correspond with siren inhabited water bodies throughout South Texas, in order to contribute to the sparse knowledge of habitat preference and siren distribution. In addition, with the growing availability and feasibility of molecular tools, I will utilize mitochondrial DNA for species delimitation and intra- and interspecific relationships. In the wake of the amphibian decline, it is pertinent to assess the population status of sirens in South Texas and determine species boundaries to provide accurate data for future management and conservation plans.

Genus Siren The published data available for genus Siren is sparse and ambiguous. At least two documented species are known to exist (S. intermedia and S. lacertina), but many aspects regarding siren life history are unknown (Petranka, 2010). Due to the relatively elusive behavior of these amphibians, data hails primarily from regions where large populations are known to occur, such as Florida and Georgia (Sorensen, 2004). Few scientific studies assessing siren populations have been conducted in Texas, and none have been conducted in the Rio Grande Valley of South Texas (Gehlbach and Kennedy,

1978; Godley, 1983; Hampton, 2009).

Habitat Very little is known about siren habitat preference, and the scarce preexisting information is debatable. Sirens are an obligate aquatic species, only traveling over land to move from one body of water to another, typically during flooding events, and have been documented in both permanent and ephemeral bodies of water (Petranka, 2010;

Schalk and Luhring, 2010; Schalk and Luhring, 2010; Tipton et al., 2012). Most often, sirens are reported in freshwater regions that are semi-shallow and stationary, such as ponds, wetlands, resacas, and even drainage ditches (Gehlbach and Kennedy, 1978;

Tipton et al., 2012). Some studies suggest sirens are most often collected in shallow water bodies with dense vegetation, but details regarding the plant communities are not documented despite its potential importance in habitat preference (Davic and Jr., 2004;

Gehlbach and Kennedy, 1978; Schalk and Luhring, 2010). In addition, sirens are collected in water bodies with rich sediments, and abundant mollusk communities

(Gehlbach and Kennedy, 1978).

Siren habitat preference may be driven by biotic factors such as predator prey interactions and available food resources. Most salamanders are considered dominant vertebrate predators, controlling the density of species in lower trophic levels (Davic and

Jr., 2004; Gehlbach and Kennedy, 1978). In East Texas, S. intermedia were documented as contributing 38-57 g/ m2 of the total standing crop biomass, and estimated densities have ranged from 0.33 sirens/m2 to 1.3 sirens/m2 (Gehlbach and Kennedy, 1978;

Hampton, 2009). Similarly, robust densities of S. lacertina were documented in Florida with 1.3 sirens/m2, suggesting sirens comprise a large portion of the biomass of their habitat, and likely play a prominent role in the food web of these ecosystems (Sorensen,

2004).

Despite high densities of sirens, co-occurring species within a water body can affect siren presence through direct resource competition or predatory consumption

(Snodgrass et al., 1999). Sirens are a highly productive species that exhibit high fecundity and rapid growth, and may be restricted to water bodies with abundant prey items or few

predators (Gehlbach and Kennedy, 1978). Several studies suggest sirens feed most heavily on mollusks, small vertebrates, worms, and aquatic insects (Conant and Collins,

1998; Hanlin, 1978; Hill et al., 2015; Petranka, 2010; Schalk et al., 2010), but a recent study from Hill et al. (2015) reported direct observation of facultative herbivory among two species of Sirenidae, and suggested this behavior may be observed throughout this family. Thus, in contrast to many salamanders, sirens appear to be omnivorous rather than solely carnivorous (Petranka, 2010). Natural predators of sirens include large wading birds, snakes, and even large fish (Tipton et al., 2012). In Florida, American alligators regularly consume S. lacertina (Delany and Abercrombie, 1986). The Western

Mud Snake (Farancia abacura reinwardti) and the Mississippi Green Water Snake

(Nerodia cyclopion) have also been reported to feed on sirens, as well as amphiumas

(Werler and Dixon, 2000).

Abiotic factors, such as water temperature, pH, depth and water body permanence also affect species diversity and richness of aquatic habitats (Dodd and Smith, 2003). We hypothesize that, as an obligate aquatic amphibian species, siren distribution may also be restricted by these abiotic factors. Though general characteristics of siren habitat have been documented, the minute details of siren habitat preference, with regards to environmental variables, are unclear (Petranka, 2010; Tipton et al., 2012). Little is known about the favorable pH and conductivity conditions for sirens, but studies have documented sirens in water bodies with low dissolved oxygen levels (hypoxic), becoming depleted as water temperatures increase in the summer (Duke and Ultsch, 1990;

Snodgrass et al., 1999; Tipton et al., 2012). Studies suggest the external gills in

conjunction with the lungs enable sirens to survive in extremely hypoxic conditions

(Tipton et al., 2012).

Studying abiotic and biotic factors is important for any habitat analysis, but is even more critical in aquatic habitats where ecosystems are extremely sensitive, and can fluctuate drastically, such as resacas (Dodd and Smith, 2003). Both biotic factors such as vegetation, food supply and species richness, and abiotic factors, such as water temperature and pH, likely play defining roles in habitat preference, and ultimately species distribution of the South Texas siren.

Seasonality Seasonality impacts siren movement and dispersal, and may differ between siren species and geographic location (Ultsch, 1973). Seasonality refers to the physiological responses that result from fluctuating environmental conditions, largely influenced by time of year (Chan, 2003). Siren seasonality has been documented in various geographic locations in the United States, but has not been studied in South Texas, which experiences a spectrum of environmental extremes (Raymond, 1991).

Seasonality may affect reproduction, abundance, and activity levels of sirens

(Chan, 2003; Raymond, 1991). Petranka (2010) suggests that sirens are most active during the summer months (June and July) when water temperatures are warmer and food sources tend to be more readily available, and are least active during winter months

(December to February) when water temperatures are low. Gehlbach and Kennedy (1978) showed that S. intermedia in East Texas displays no difference in seasonal activity levels, whereas Hampton (2009) suggested the same species to be more active in late winter and early spring (Collette and Gehlbach, 1961; Hampton, 2009). In Louisiana, S. intermedia

seemed to show greater activity levels during the fall and winter (Raymond, 1991).

Hanlin and Mount (1978) suggested that S. lacertina, was the most active during summer in Alabama, and Sorensen (2004) suggested winter in Florida. For both S. lacertina and

S. intermedia, breeding is seasonal, and occurs between the late winter and early spring

(November to March), which seemingly coincides with greater activity levels (Collette and Gehlbach, 1961; Sorensen, 2004; Tipton et al., 2012). This timeframe is suggested based on the presence of bite scars on females (a supposed mating ritual) and from an increase in juvenile individuals in the following season (Fauth and Resetarits Jr., 1999;

Godley, 1983; Hampton, 2009; Tipton et al., 2012).

Seasonality induces a unique survival technique for sirens called aestivation.

Aestivation is a response to drought in which the sirens burrow into the mud and remain in a state of dormancy until optimal conditions and water return (Conant and Collins,

1998). This technique has been observed in other species, such as the African lungfish, yet the trigger for this induced state is unclear (Fishman et al., 1992). Aestivation enables siren survival, but depending on drought duration or recurrent seasonal drought, may impact siren abundance (Gehlbach et al., 1973). Larger siren individuals have a greater percentage of body mass to lose, and can thus survive longer amounts of time in aestivation than juveniles (Gehlbach et al., 1973). In a laboratory experiment, one S. lacertina individual aestivated for 5.2 years, in which time 86% of the initial body weight was lost (Etheridge, 1990). Aestivation is advantageous for sirens because it allows survival in regions restricted to other aquatic organisms, which likely facilitates siren dominance, and affects siren distribution (Gehlbach et al., 1973).

Distribution Sirens are found only in North America (Petranka, 2010). There are currently two documented species of siren in the United States, S. lacertina and S. intermedia, yet the distribution of these species remains blurred (Parra-Olea et al., 2008). Siren lacertina occurs in permanent and semi-permanent water bodies extending from Washington, D.C. to southern Florida, to Alabama (Petranka, 2010). The majority of the S. lacertina population lies in the coastal plains, with population abundance in Florida and Georgia.

Siren intermedia has a more expansive range than S. lacertina, and can be found along the eastern coast of the United States extending into Florida, and throughout east Texas extending to southern Michigan (Petranka, 2010).

Though northeast Texas contains abundant populations of S. intermedia, South

Texas is an area of contention regarding current siren distribution. Sirens are found in areas of Northern Tamaulipas, Mexico and the Rio Grande Valley of Texas, as far as

Webb and Nueces counties (Conant and Collins, 1998; Tipton et al., 2012). Published studies based on museum voucher specimens suggest that both S. lacertina and S. intermedia could be sympatric in South Texas and Northern Mexico (Flores-Villela and

Brandon 1992). Siren lacertina has historically been documented in Cameron, Duvall,

Victoria, and Maverick Counties, while S. intermedia has been recorded in Victoria and

Cameron counties in Texas, as well as Tamaulipas, Mexico (Brown, 1950; Oscar Flores

Villela and Ronald A. Brandon, 1992). The legitimacy of some of these historical siren records is questionable, yet the records suggest the South Texas siren could be composed of S. intermedia individuals, S. lacertina individuals, a combination of both, or an endemic taxon. Most recently, Tipton et al. (2012) have suggested that the sirens of South

Texas are a distinct species (referred to as the “Rio Grande siren”), but a robust approach is needed to validate this hypothesis (Tipton et al., 2012). Thus, the distribution of the

South Texas siren is unclear because the species of the population is unknown.

Species identification Two studies have attempted to resolve the taxonomic status of sirens in South

Texas based solely on morphological characteristics; however, the species identification of the South Texas siren remains unresolved (Goin, 1957; Oscar Flores Villela and

Ronald A. Brandon, 1992). Discrepancies in the literature, historical documentation, and the identification of museum vouchers, provide an inaccurate and muddled dataset, primarily because visual identification of sirens is difficult.

Historically, salamanders have been classified to species by counting the number of costal grooves, vertical grooves that extend along the sides of many amphibians

(Lopez and Brodie, 1977). Siren lacertina has been recorded as having 36 to 40 costal grooves, most often with 37-38 (Chauncey Bishop, 1943; Conant and Collins, 1998;

Petranka, 2010). Siren intermedia has a modal number of 35 costal grooves (Chauncey

Bishop, 1943) and the South Texas siren has been documented having 36-38 costal grooves (Conant and Collins, 1998; Tipton et al., 2012). Coloration patterns and morphology have been used to identify species, but all sirens share nearly identical morphology, and color patterns can differ greatly even within a species. Both S. intermedia and the South Texas siren are supposedly differentiated from S. lacertina by having pointy tails (Tipton et al., 2012). All sirens (S. lacertina, S. intermedia, and the

South Texas siren) are brown, grey, olive green, or a bluish color, with varying degrees

of black spots and light blotches along the body (Petranka, 2010). Size has shown to be an effective tool for species identification of mature individuals because S. lacertina is the largest of the species (≤ 980 mm), but juvenile S. lacertina are nearly indistinguishable from mature S. intermedia (≤ 500 mm) (Chauncey Bishop, 1943;

Petranka, 2010). Sirens in South Texas have been documented as large as 690 mm, further confusing species identification (Petranka, 2010; Tipton et al., 2012).

Goin (1957) and Villela and Brandon (1992) used size, coloration, and costal groove number to identify sirens in South Texas, yet due to an overlap in body length and costal groove number, two taxa were recognized (Goin, 1957; Oscar Flores Villela and

Ronald A. Brandon, 1992). Goin (1957) proposed the sirens in South Texas are a subspecies of S.intermedia, based on phenotypic variation, and referred to it as the “Rio

Grande Siren.” In contrast, Villela and Brandon (1992) declared the South Texas siren to be S. lacertina. Neither study utilized molecular techniques for species identification.

Similarities in coloration, morphology, and size overlap render visual identification insufficient as a sole means of species classification, thus species identification of the

South Texas siren will likely be determined through genetic analysis.

Mitochondrial DNA Many salamander species that are morphologically similar are identified through genetic analysis (Funk et al., 2012; Vences et al., 2005b). Many amphibians, including salamanders, have large genomes compared to other vertebrate species, which provides a complex yet useful template for genetically identifying species and analyzing populations and their evolutionary history (Steinfartz et al., 2004). While various ways to genetically identify species have been developed, the use of molecular markers has proven to be a

reliable and commonly used tool (Vences et al., 2005b). Species identification has been employed using both nuclear DNA and mitochondrial DNA, but mitochondrial DNA is advantageous (Rubinoff et al., 2015; Weisrock et al., 2005).

Mitochondrial DNA (mtDNA) markers are often utilized to differentiate species, determine evolutionary relationships, and examine relatedness among populations (Arif and Khan, 2009; Vences et al., 2005b). Mitochondrial markers are a useful means of genetic analysis because these genes are conserved through generations and provide direct connection through maternal inheritance (Arif and Khan, 2009). For the complete mitochondrial genome, genes are encoded on both the heavy strand (leading) and the light strand (lagging) (Samuels et al., 2005). There are typically 2 rRNA subunits, 13 protein coding genes, 22 tRNAs, a non-coding region (D-Loop) an intergenic spacer

(IGS), and an origin of light strand replication (OL) within the amphibian mitochondrial genome (Samuels et al., 2005; Zhang et al., 2008).

Single mtDNA genes are utilized for species identification and for population studies (Kuchta and Tan, 2004; Rubinoff et al., 2015; Vences et al., 2005b). The most commonly used mitochondrial gene regions for species identification are the ribosomal subunit 16S, the cytochrome-oxidase 1 protein coding gene (CO1), and the protein coding gene cytochrome-b (Cyt-b) (Vences et al., 2005b). 16S is a highly conserved ribosomal subunit and can be used to designate phylogenetic inferences as broad as phyla (Arif and Khan, 2009). Despite being highly conserved, 16S does show variability in areas of nucleotide mutation that may be species specific (Guha et al., 2006; Vences et al., 2005a). Thus, 16S has been proposed as an ideal molecular marker for amphibian species identification, due to a sufficient amount of mutations for lower taxonomic

resolution (Arif and Khan, 2009; Grosjean et al., 2015; Vences et al., 2005b). For example, the 16S gene successfully deciphered between two anuran species and their hybrid cross (Lamb et al., 2000). The protein-coding gene CO1 is the most utilized molecular marker across the animal kingdom, and has been termed the “barcode of life”

(Arif and Khan, 2009; Ratnasingham and Hebert, 2007; Vences et al., 2005b). Protein coding genes, such as CO1, may be more powerful molecular markers than ribosomal mitochondrial markers because they are still highly conserved but evolve at a greater rate, which can assist with more recent genetic divergence (Arif and Khan, 2009). Despite the utility of CO1, this gene has been used cautiously with amphibian studies due to its high variability and has been credited as misleading for amphibians because of its role in the explosion of “new” amphibian species (Mueller et al., 2004; Smith et al., 2008; Vences et al., 2005b). Cyt-b is another extensively utilized gene because of the availability of universal markers and is primarily used for amphibian phylogenetic analyses (Guha et al.,

2006; Matsui et al., 2007; Mueller et al., 2004). For species identification, Cyt-b has been problematic because of the lack of variation to taxonomically identify below the genus level (Branicki et al., 2003). Non-coding regions, such as the intergenic spacer (IGS) or

D-Loop, may be useful for identifying recent evolutionary divergences, but may be too variable for species identification (Lunt et al., 1998). The rates of evolution among these genes are not constant, but are influenced by evolutionary pressures such as genetic drift and restricted gene flow, and the physical location within the genome (Helm-Bychowski et al., 1985; Rubinoff et al., 2006). Based on the asymmetrical replication of the mitochondrial genome, gene regions further away from the origin of replication accumulate mutations faster (Gibson, 2004). These evolutionary constraints can affect the

degree of gene utility and accuracy, and may yield differing species identification results between genes (Helm-Bychowski et al., 1985; Rubinoff et al., 2006). Thus, larger regions of the mitochondrial genome that can encompass a greater degree of the variation may more accurately identify species.

While individual genes are often employed to uncover population genetics and identify species, the use of the complete mitochondrial genome (CMG) may be a more accurate and robust mechanism for species identification (Cummings et al., 1995; Jiang et al., 2013; Zhang and Wake, 2009; Zhang et al., 2008). This is especially true due to the variability between evolutionary rates and nucleotide substitutions of single genes

(Rubinoff et al., 2006; Vences et al., 2005b). In recent years, sequencing the CMG has become more prevalent for phylogenetic analyses because it provides more accurate estimations of evolutionary relationships, better resolution for deeply divergent lineages, and increases accuracy due to the robust amount of mtDNA available for analysis (Jiang et al., 2013; Zhang and Wake, 2009; Zhang et al., 2008). Numerous studies across a number of vertebrates have adopted the use of the CMG for clarity in phylogenetic relationships (Samuels et al., 2005; Yu et al., 2007). Zhang et al (2009) sequenced the

CMG for all genera (not previously sequenced) within the family Salamandridae, and has seemingly resolved the true phylogenetic relationships among salamanders, where previous studies using mitochondrial fragments yielded discordant phylogenies (Zhang et al., 2008). Though the CMG has deciphered more accurate phylogenetic relationships than single genes, surprisingly, very little research has been conducted employing the use of the CMG to delimitate species on a fine scale level, especially for amphibians.

Often times, species identification requires a large data set of representatives from varying localities, which can make sequencing the entire genome expensive and time consuming (Mueller et al., 2004). In the wake of this dilemma, individual genes have provided enough resolution for some species identification, but not in the most encompassing and resolute manner (Rubinoff et al., 2006). By sequencing the complete mitochondrial genome for species identification, we will accurately assess individual genes for specific species, and identify the most useful mitochondrial gene for future species identification and phylogenetic inference (Mueller et al., 2004).

Objectives and Hypotheses The aim of this study was threefold: to resolve the taxonomic identity of sirens in

South Texas using mtDNA, to determine Siren species distribution in South Texas, and to assess the characteristics of sampled water bodies for correlation with siren presence. For species identification, the major objective was to determine if siren species could be determined by a single mtDNA gene, or if the complete mitochondrial genome was necessary to resolve taxonomic identity. To understand Siren distribution, the objective was to use the mtDNA sequences of Siren from South Texas to resolve species ranges, and to determine genetic connectivity among sirens throughout South Texas. For Siren habitat assessment, the overarching goal was to identify potential correlations of siren presence within a water body by characterizing the vegetation, co-occurring species, and environmental variables. To meet these goals, the following hypotheses were tested:

(1) Water bodies with siren presence will differ in vegetation composition, co- occurring species composition, and environmental characteristics to water bodies where sirens are not found because Siren distribution is determined by community structure and abiotic factors.

(2) Sirens endemic to South Texas consist of a single species, and will be genetically distinct from published sequences for Siren lacertina and Siren intermedia based on geographic isolation.

(3) South Texas siren species identification will be better resolved with the complete mitochondrial genome sequence than with single mtDNA genes because the CMG is more robust and accounts for greater variation, but the 16S and CO1 genes will identify sirens as the same species.

(4) Siren population structure will vary regionally, with sirens in adjacent water bodies being more genetically similar than those in geographically distant locations, based on the limited dispersal abilities of Sirens.

II. MATERIALS AND METHODS

Study Sites Field collections for this study took place in Hidalgo, Cameron, and Kleberg

Counties (Figure 1). Cameron County and Hidalgo County are the two southernmost counties in Texas, and are bordered by the Rio Grande River to the South. These two counties are bordered to the North by the Coastal Sand Plain of Texas, separating them from Kleberg County. Water bodies within this region were selected for sampling based on water availability, documented Siren spp. occurrence, and accessibility. Site accessibility was dependent on Scientific Research permits from Texas Parks and

Wildlife, federal Special Use permits from U.S. Fish and Wildlife Service, and permission from private landowners. Sampled water bodies included resacas, drainage ditches, ponds and ephemeral water bodies, within the South Texas Refuge Complex, wildlife management areas, private properties, and wildlife sanctuaries.

Trap Collections and Environmental Variables To assess the presence of sirens, water bodies were sampled on at least one occasion. Initially, all sites were to be sampled twice, but only some were repeatedly sampled due to low water levels in the summer. Vinyl-dipped, 16½”, 2-piece Minnow

Traps (Academy Sports and Outdoors, Katy, TX), with funnel ends manually widened, were used for capture. All traps (after 9/28/2013) contained a flotation device (foam water noodles). Traps were baited with small, plastic containers with holes, filled with either chicken gizzards or chicken necks. In 2013, traps were placed at approximately 6- meter intervals. In 2014, trap placement was further standardized for ease and for a more systematic sampling technique. Five traps were connected to a single line of rope

(referred to as a “trap line” moving forward) in pre-measured increments of 6 meters.

GPS coordinates were collected for every individual trap using a GARMIN eTrex 10

(Garmin Corp., Olathe KS). Traps were set along the edge of the water body, at an average water depth of 27.2 cm, though water depth ranged from 10 to 98 cm.

Upon deployment of traps, pH, water temperature, conductivity, depth, and dissolved oxygen readings were collected at the first, third, and fifth traps on the trap line.

Repeated samples were collected along a trap line to account for single measurement variability and were averaged within a site for analysis. Temperature and pH were collected using a Waterproof Digital pH Meter Tester Thermometer °C/°F ATC

Electrode Dual Display (China). Dissolved oxygen was determined in mg/L using an AZ

8403 Dissolved Oxygen Meter (AZ Instrument Corp., Taiwan) and conductivity in µs/cm using a 138 (II) Conductivity Tester (Kelilong Electron Co., China). Depth was measured with a meter stick. Devices were calibrated before sampling.

Vegetation within the water bodies was identified by structural group, or vegetation type (Table S1) (Schalk et al., 2010). Plant life was assessed along the edge of the water body (edge vegetation), within the water body but emerging from the water

(emergent vegetation), and within the water body, but submerged (submerged vegetation)

(Figure S1) (Peterson and VanderKooy, 1997). Vegetation type was solely identified as present or absent within the three described zones. Upon collection of environmental variables, deployed traps were left on site overnight, for approximately 18 hrs, and retrieved the following morning.

Upon retrieval, all traps were evaluated for siren presence and abundance. All co- occurring species collected in traps were documented for abundance, and identified to the

lowest taxonomic level (Table S2). All captured sirens were transferred from the minnow traps to five gallon buckets containing water from the capture site. First, sirens were weighed to the nearest 0.1 gram in a small, tared container. Each siren was then measured for total length (mm) in a long, rectangular container using a sewing tape measure. For more precise readings, the sirens were pushed up against the long side of the container with a plexiglass rectangle (Figure S2). The weighing tray, containers, and any other materials used were rinsed and wiped clean between individuals using water from the collection site. Morphological characteristics were documented, but were difficult to articulate. Many attempts were made to count the costal grooves along the sides of the sirens, but due to the erratic and non-stationary behavior of the sirens, even when held in tubes, only a few sirens were counted and subsequent costal groove counts were abandoned.

Following morphometric measurements, all sirens were photographed.

Photographic identification was used to identify all sampled siren individuals. Initial siren photographs were captured on a black background using an iPhone 4S camera (Apple

Inc., Cupertino, CA), but were subsequently taken on a white background with a Canon

Rebel Ti3 (Canon USA, Melville, NY) with a macro lens and flash lights. Siren individuals were submerged in water for the photographs to reduce glare. Siren individuals were not restrained or anesthetized for handheld photographs. The dorsal region of the siren head was of primary interest for photographs for individual identification using Wild-ID, as has been done by other studies identification (Bendik et al., 2013).

In addition to sampling single sites for siren presence, a consecutive sampling period was conducted at a single pond location that was known to contain sirens. Five minnow traps were set along the edge of a single water body, with approximately 6 m between traps. Traps were initially set on July 13, 2014, were checked for siren presence, and were re-baited every morning until August 9, 2014. Traps were only assessed for siren presence and abundance. The aim of this trapping experiment was to determine the catch per unit effort (CPUE) for sirens in water bodies in South Texas during the peak summer season.

Photo-ID Wild-ID is an open-source pattern identification software that uses the SIFT

(Scale Invariant Feature Transform) algorithm to assess differentiable patterns within photographs, and to compute matching scores from photograph comparisons within the database (Bolger et al., 2012). Wild-ID has been used to differentiate between individuals of giraffe (Giraffa camelopardis), wildebeest (Connochaetes taurinus), and even the small, Jollyville Plateau salamander (Eurycea tonkawae) (Bendik et al., 2013; Bolger et al., 2012; Morrison and Bolger, 2012). The SIFT software focuses on key attributes and features that characterize the patterns of interest. Keypoints are identified within the patterns and are compared between photographs of similar orientation and scale using a goodness-of-fit model. Pattern-matching numerical scores are computed for every photograph comparison. Photograph similarities are scored on a scale from 0.0 to 1.0, with 0.0 representing a 0% match, and 1.0 representing a 100% match. Based on previous studies that have utilized the Wild-ID program, a score of 0.1 (10%) or higher in accordance with manual screening computed a match (Bendik et al., 2013).

To validate the Wild-ID program for sirens, a control experiment was also conducted using a single Siren intermedia individual from the Gladys Porter Zoo in

Brownsville, TX. This control was set up to ensure that the same individual would be recognized using Wild-ID over an extended time period between photographs.

Sample Collection After all photographs and morphometric measurements were collected, tissue for

DNA sampling was collected from sirens on site. Sirens were handled briefly as described in (Luhring, 2008) and a small “v-notch” (roughly 5 mm by 2 mm tissue sample) was cut from the dorsal region on the end of the caudal fin using alcohol- sterilized scissors. All captured sirens were released on site, with the exception of Siren

S7, which was dead upon trap retrieval. In addition, Siren S31 was found dead near a water body. Problems sequencing the tail section DNA for S31 resulted in liver tissue

DNA being used as the template. Siren tissue samples were placed in 500 µl of ethyl alcohol (95%) in 1.5 mL DNase/RNase-free microcentrifuge tubes, and stored in a -20°C freezer within 5 hours of collection.

Habitat Analysis

Co-occurring Species Composition To determine if siren presence/absence and abundance correlated with co- occurring species, a RELATE analysis was conducted. Prior to analysis, co-occurring species composition abundances were square-root transformed. Bray-Curtis similarity matrices were produced for species compositions, siren presence/absence, and siren abundance. Non-metric multidimensional scaling (MDS) was used to visualize differences in co-occurring species abundances for each site by region. Two sites were

excluded from analysis because they were the only site within a region. Individual sample sites were categorized by region based on geographic proximity (9 regions total, Figure

1). An Analysis of Similarity (ANOSIM) test was applied to determine if species composition differed significantly by region, and a Similarity Percentages Test

(SIMPER) was applied to assess which co-occurring species contributed to region similarities.

Environmental Variables

Environmental variables (pH, conductivity, temperature) were first analyzed with a Draftsman’s plot, and were normalized for analysis. Environmental variables at each site were compared to Bray-Curtis similarity matrices for siren presence/absence and abundance using the BEST procedure.

Vegetation Composition Vegetation percent cover and vegetation group presence/absence were analyzed across three zones within all thirty-six sites. A Bray-Curtis similarity matrix was generated to compare percent cover across all three zones in all sites. An ANOSIM test was applied to determine if percent cover for all three zones together, and for all three zones separately, differed significantly between sites with siren presence/absence or abundance. A Two-Way ANOSIM test was also applied to determine if percent cover differed significantly between sites with siren presence/absence and between regions. A

MDS plot was produced to visualize percent cover composition with siren presence/absence, abundance, and by region. A SIMPER test was run to assess which zone contributed the most to similarities.

All three zones were analyzed independently for the presence/absence of nine vegetation groups and siren presence/absence. A Bray-Curtis similarity matrix was created for each zone, and an ANOSIM was conducted to determine whether there was a significant difference between site siren presence and the presence/absence of vegetation groups within each zone. An ANOSIM was also conducted for analysis of vegetation groups across regions. A MDS plot was created to visualize siren presence/absence with zone vegetation groups. A SIMPER test was run to assess which zone contributed the most to similarities. A Two-Way ANOSIM test was applied to determine if there was a significant difference between the vegetation groups in the three zones across all sites with the presence/absence of sirens. Statistical significance was determined at p < 0.05 for all tests used in this study. All multivariate data analyses of habitat characteristics were conducted in PRIMER-E v6 software. mtDNA Analysis

DNA extraction, cloning, and sequencing Total DNA was extracted from tissue samples of 22 siren specimens using the

GenCatch™ Blood &Tissue Genomic Mini-Prep Kit (Epoch Life Science Inc., Sugarland

TX). The 22 specimen included 18 individuals from the unidentified siren populations of

South Texas (South Texas siren), two donated samples from the Attwater Prairie Chicken

NWR in Texas, one specimen from Florida identified as Siren lacertina by Paul Moler, and one unidentified siren voucher specimen (denoted as siren SG) collected from

Williamson county, TX (on loan from the Texas Natural History Collections) (Figure 2).

Extracted DNA was used as a template for Polymerase Chain Reaction (PCR).

Three fragments of Mitochondrial DNA of the expected sizes of 5693 Kb, 5733 Kb and

6887 Kb were amplified with siren-specific primer sets (Table 1). PCR reactions were conducted with GoTaq Long PCR Master Mix (Promega Corp., Madison WI) in total volumes of 50 µl. Each reaction included 50-100 ng of total DNA. The amplification conditions for fragments L1-C2 and C2-G are as follows: hot start at 95°C for 2 min; 35 cycles of denaturing at 94°C for 30 sec, annealing at 64°C for 30 sec, and extension at

65° for 6 min; and a final ending step of 72°C for 10 min with a 4°C hold for 30 sec. The amplifications for G-L1 were the same, except the annealing temperature was 58°C, the extension step was 7 min, and 37 cycles were run. PCR products were visualized using

Gel Electrophoresis (0.7% TBE agarose) with an LED transilluminator, and bands were cut and purified using the GenCatch™ Advanced Gel Extraction Kit (Epoch). Gel- purified PCR products were ligated into the pGEM®-T Vector as described by the manufacturer’s instructions (Promega). Transformations of competent Escherichia coli

(DH5α) were performed by heat shock (Maps). Transformants were plated on LB-agar

100ug/ml ampicillin and clones were screened by PCR with T7 and SP6 universal primers. One recombinant plasmid harboring the expected product size was selected for sequencing for each fragment per siren.

A total of twenty-one primers were used to primer walk the three cloned fragments to sequence the entire mitochondrial genome (Figure 3, Table 1). Ten degenerate salamander primers (Zhang et al., 2008) and eleven siren-specific primers were used to sequence the cloned regions, in addition to the T7 and SP6 primers. Six primers were used to sequence Fragment L1-C2, seven primers were used for Fragment

C2-G, and eight primers were used for Fragment G-L1 (Figure 3). Due to genetic variation across sampled sirens, some primers failed to successfully amplify products for

all sirens. Thus, alternate primers were developed when sequencing reactions failed

(Table 1).

Sequence Assembly Successive sequences were aligned with Clustal Omega (EMBL-EBI, Hinxton,

Cambridge, United Kingdom) to identify overlapping sequence regions for assembly.

Overlapping segments were visualized for nucleotide discrepancies using CLC Main

Workbench 7.6.1 (CLC Bio, Aarhus, Denmark). Sequence discrepancies were manually adjusted with data from sequence chromatogram files. In cases where nucleotide discrepancies were not resolved with chromatogram data, a majority rule consensus call was made based on siren-only sequence alignments. Sequence alignment was performed using the MUSCLE (multiple sequence comparison by log-expectation) algorithm with default parameter settings in CLC Main Workbench 7.6.1 (CLC Bio, Aarhus, Denmark) and MEGA6 (Edgar, 2004; Sievers and Higgins, 2014).

This process allowed the entire mitochondrial genome to be sequenced for nine individual sirens, including 8 sirens from the unidentified siren populations of South

Texas (South Texas siren) and one S. lacertina specimen from Florida. The entire mitochondrial genome was not sequenced for all sirens. Instead, partial mitochondrial genome sequences were generated from the remaining sirens to provide additional data for independent analysis of protein coding genes and non-coding regions. The partial genome sequences consisted of cloning the single fragment regions of the genome (L1-

C2, C2-G, or G-L1) (Figure 4). Three L1-C2 and C2-G fragment sequences, and nine G-

L1 fragment sequences were generated from the additional sirens.

Sequence Alignments Phylogeny for Order Caudata was investigated by aligning all 9 siren CMG sequences from this study with 5 amphibian CMG sequences retrieved from GenBank.

The amphibian genomes used for analysis included four species from order Caudata

(Siren intermedia, GenBank no. GQ368661; Pseudobranchus axanthus, GenBank no.

GQ368660; Ambystoma tigrinum tigrinum, GenBank no. AY659992; and Dicamptodon aterrimus, GenBank no. GQ368657) and one out-group from Order Gymnophiona

(Typhlonectes natans, GenBank no. NC002471). Alignments were constructed in CLC

Main Workbench 7.6.1. and MEGA6 using the MUSCLE alignment default parameter settings (Tamura et al., 2013). Siren sequences from this study were annotated for tRNAs, ribosomal subunits, protein coding regions, and non-coding regions based on the alignment consensus. Siren protein-coding nucleotide sequences were translated into amino-acid sequences, and discrepancies in nucleotide translations were manually adjusted. The amphibian sequences, S. intermedia and D. atterimus, contained missing data (denoted by “N”) in tRNAARG, but were used in analysis. Due to length variation and unidentifiable regions, the following regions were excluded from phylogenetic analysis: tRNATHR, tRNAPRO, the intergenic spacer (IGS), and the non-coding control region (D-

Loop). The origin of replication sequence between tRNAASN and tRNACYS was also excluded from analysis. The siren+outgroup data set was composed of 20 tRNAs, 2 rRNA subunits (12S, 16S), and 13 protein coding genes (ND1, ND2, CO1, CO2, ATP8,

ATP6, CO3, ND3, ND4L, ND4, ND5, ND6, Cyt-b) for a total of 15541 base pairs.

In addition to the alignment described above, a siren-only CMGS alignment was also constructed. This alignment included only the 9 siren CMGS from this study. Eight

of the sirens sequenced were from Texas, and one siren was a S. lacertina from Florida.

This alignment contained all tRNA’s, ribosomal units, protein coding genes, the intergenic spacer (IGS), and the non-coding D-Loop region. The origin of replication sequence between tRNAASN and tRNACYS was excluded. The siren-only dataset was composed of 22 tRNAs, 2 rRNA subunits (12S, 16S), 13 protein coding genes (ND1,

ND2, CO1, CO2, ATP8, ATP6, CO3, ND3, ND4L, ND4, ND5, ND6, Cyt-b), one intergenic spacer (IGS), and the non-coding control region (D-Loop), for a total of 17149 base pairs.

Sequence alignments were also constructed for the 16S rRNA gene, the protein coding regions, cytochrome oxidase subunit 1 (CO1) and cytochrome b gene (Cyt-b), and the intergenic spacer for independent analysis. S. intermedia and P. axanthus GenBank sequences were utilized as out-groups for all single gene sequence alignments with all available siren sequences for that particular gene. The 16S, CO1, Cyt-b, and IGS alignments included 13, 11, 18, and 18 sirens from this study, respectively. The intergenic spacer region was not sequenced for the S. intermedia and P. axanthus

GenBank sequences, excluding them from analysis. Aligned sequences for the 16S, CO1,

Cyt-b and IGS gene analysis, were 1604 bp, 1554 bp, 1141 bp, and 850 bp long, respectively.

Phylogenetic Analysis

Bayesian Analyses Maximum Likelihood (ML) and Bayesian phylogenetic analyses were conducted separately on all data sets to analyze phylogeny. The Bayesian phylogenetic analysis was implemented using MrBayes version 3.2 (Ronquist and Huelsenbeck, 2003). For

Bayesian analysis, the siren-outgroup dataset and the siren-only datasets were partitioned.

All tRNAs were concatenated into a single partition (Ronquist and Huelsenbeck, 2003).

Each individual ribosomal subunit (2) and protein coding genes (13) were treated as separate data partitions. For the siren-only dataset, the non-coding intergenic spacer and the D-Loop were also treated as separate data partitions. Model selection for each partition within the dataset was chosen for all according to the Akaike information criterion (AIC), as implemented by PartitionFinder (Lanfear et al., 2014). All three codon positions of protein coding genes were analyzed for separate nucleotide evolutionary models using a heuristic search algorithm, and revealed that models of evolution varied within genes (Lanfear et al., 2014). The models of evolution also took into account gamma distributed rate variation among sites (G) and the proportion of invariable sites

(I). Thus, Partition Finder determined the best-fitting partitioning scheme to contain 19 subsets for the siren+outgroup dataset, with a total of 5 different models (Table S3). For the siren-only dataset, 17 subsets were designated, with a total of 8 different models

(Table S3). For both datasets, the predominant model was General Time Reversible

(GTR).

Bayesian analysis was run for 1 million generations, and was sampled every 1000 generations, with 25% of the generations discarded as burn-ins. Four Markov chain runs were utilized, and partition sets were unlinked to account for nucleotide substitution variation across genes. The analysis was repeated twice for each data set to assess the robustness of the posterior probabilities. The posterior probabilities (PP) are the probabilities that the phylogenetic tree is correct, and ranges from 0 to 1, with 1 signifying absolute resolution.

Maximum Likelihood Analyses Data sets were also analyzed in MEGA 6 for a Maximum Likelihood analysis

(Tamura et al., 2013). Treating each data set as a single partition, the siren-only alignment, the siren+outgroup alignment, 16S, CO1, Cyt-b, and the IGS were analyzed for a best-model in MEGA 6. MEGA 6 determined GTR+G to be the best model test for

Maximum Likelihood analysis for all alignments.

For tree construction, gaps were initially tested in three different manners: complete deletion, use all sites, and partial deletion at 75%. Phylogenetic trees showed minimum variation among the three treatments, thus ‘use all sites’ was applied for analysis. This is likely the best option because minute differences within the siren-only alignments may prove substantial for accurate species analysis. ‘Use all gaps’ was applied to all datasets. To assess branch support, bootstrap values were set to 1000 replicates with a strong branch swap filter. The bootstrap values represent the overall strength of the phylogenetic tree, and ranges from 0 to 100.

Sequence Divergence Analyses Sequence divergence and nucleotide diversity of sirens was examined for the large siren+outgroup data set and the siren-only data set using the Kimura 2-parameter corrected pairwise genetic distance model, calculated in MEGA 6 (Tamura et al., 2013).

The Kimura 2-parameter genetic distances were used to assess mtDNA sequence divergence by taking both transitional and transversional substitution rates into account, while assuming equal rates of nucleotide substitution among sites (Kimura, 1980; Vences et al., 2005b). Because all sites could not be used in MEGA 6, analyses were conducted with a partial deletion set at the lowest percent (5%) to maintain gaps. For each gene, all

available siren sequences from this study and the S. intermedia sequence from GenBank were analyzed; however, P. axanthus was included in the sequence alignment solely for inclusion in the topology of the phylogenetic tree. Genetic distances were also analyzed for the 16S, CO1, Cyt-b, and IGS to test for phylogenetic utility.

To assess whether there is a relationship between genetic distance and geographic distance for the sirens in this study from Texas, estimates of pairwise uncorrected p- distances were determined for 16S, CO1, Cyt-b, and IGS by MEGA 6 (Tamura et al.,

2013). The uncorrected p-distances were calculated as the number of base differences per site between sequences. Uncorrected p-distance values are most commonly used for comparison of mean pairwise genetic distances between clades or populations for species assessment, but due to small sample size, raw values were compared between individual siren sequences (Fouquet et al., 2007; Grosjean et al., 2015; Kuchta and Tan, 2004;

MartíNez-Solano et al., 2007; Matsui et al., 2007). To determine whether genetic variation among sirens is caused by distance isolation (IBD), uncorrected p-distances and straight-line geographic distance (km) were tested for correlation using full Mantel tests with the IBD web service (Jensen et al., 2005). Straight-line geographic distances were calculated between samples from the collection site’s latitudinal and longitudinal GPS coordinates. For the SG siren, GPS coordinates were estimated from the collection notes description. Four analysis matrices were run for each dataset, using log (genetic similarity) and log (geographic distance) jointly and separately to determine the best-fit model (Bohonak, 2002). The Mantel test assessed the statistical significance of correlation between the pairwise genetic distance matrix and the pairwise geographic distance matrix (Bohonak, 2002). In addition, the IBD web service provided Reduced

Major Axis (RMA) regression estimated for the slope and intercept, providing a true correlation relationship (r2) for the variables (Bohonak, 2002).

III. RESULTS

Siren Collection Data Between September 2013 and March 2015, a total of 63 sirens were captured from seventeen of forty-seven sampled water bodies, within 6 regions (Figure 1).

Sampling occurred in September, October, November, and December of 2013 and May,

June, July, August and October of 2014. Sirens were captured in September, October, and

November of 2013, May, June, July, and October of 2014, and opportunistically in March of 2015. Sirens were captured in Kleberg, Cameron, and Hidalgo Counties in South

Texas (Figure 1). October of 2013 and 2014 produced the greatest abundance of sirens collected throughout the study (Figure 5). Abundance results differed greatly from the historical records of siren captures in all of Texas, and specifically South Texas (Figure 5,

Figure S3). The month of September 2013 yielded the highest catch-per-unit-effort

(CPUE) of 0.375 sirens per trap night (Figure 6). Over a period of 28 days, two sirens were caught at the single site sampling location used to assess catch per unit effort

(CPUE). This sampling period resulted in a CPUE of 0.01 sirens per trap night. Both sirens were caught in the same trap on the same day. The two sirens were the smallest measured throughout the entire study weighing 5.9 g and 4.6 g, with corresponding lengths of 126 mm and 110 mm, respectively.

All sirens exhibited the same general morphology with two forelimbs, and branched, external gills, but length, weight, and coloration patterns differed (Figure 7).

The mean siren weight was 161 g, with weights ranging from 4.6 g to 405 g (Figure 8).

The mean siren length was 360 mm, with lengths ranging from 110 mm to 535 mm

(Figure 8). Tail shape was difficult to assess; all tails tapered to the end and were semi- pointed, though the extent of the point varied (Figure 9). Smaller sirens (< 100 g) exhibited tails that were slightly rounder than the larger sirens. All tails did have a wide extension of skin above and below the main tail body. Coloration varied greatly among captured sirens as well; small sirens (< 100 g) displayed a mottled, light olive-green color. The undersides of the small sirens were a lighter yellow-green than larger ones, and two distinct yellow lines radiated from just behind the forelimbs down the sides of their bodies (Figure 10). These lines seemed to disappear as siren sizes increased. The larger sirens’ (> 100 g) body colors ranged from solid brown, to dark olive-green with varying spot patterns (Figure 7). Most sirens exhibited small, mottled dark spots along the body, but a few individuals exhibited very large and well-defined black marks (Figure

7). All sirens exhibited very small yellow spots on the dorsal head region (Figure 11).

The yellow spots radiated in a v-shape pattern from the snout until just beyond the eyes.

Spot patterns on the head were visually distinct between individuals (Wild-ID, Figure

12).

Wild-ID Wild-ID software was used to compare all photographs taken of captured sirens.

Due to poor photo quality some sirens from 2013 were excluded from the analysis. No known recaptures of siren individuals took place, based on manual photo-analysis, known capture locations, tissue samples taken and sampling timeframe; in other words, all evidence suggests that each siren was trapped and examined only once. Using the Wild-

ID software, the majority of photograph comparisons scored a 0% match, but comparisons ranged from 0.05% to 7.35% similarities (Figure 12). Thus, with visual

analysis and Wild-ID, no recaptures were observed for the sirens compared with Wild-ID

(≤ 10%). For the control experiment for Wild-ID, the initial photograph was taken on

November 6, 2014 and the second on March 6, 2015, for a total of 120 days between photographs. From within a photo-database containing 47 other siren individuals (that could not be the captive zoo siren), the two photographs of the Siren intermedia from the

Gladys Porter Zoo were recognized as a match with a value of 0.2604 (26%), validating that we did not recapture any siren individuals from field work conducted from 2013 to

2015.

Habitat Data

Co-occurring Species Composition A total of thirty-three co-occurring species were observed within thirty-six water bodies from nine different regions between May, June, July and October 2014. The species observed were from three animal Phyla: Vertebra, Mollusca, Arthropoda (Table

S2). Observed organisms included Amphibians, Reptiles, Insects, Fish, Crustaceans and

Mollusks. Species abundances were dominated by Phylum Arthropoda (n = 2033), primarily by the predaceous diving beetle (Dytiscidae sp.), but species richness was dominated by Phylum Vertebra (n = 20) with twenty different species collected. Twenty- two of the observed thirty-three species were collected in water bodies that contained sirens. The RELATE analysis revealed that there was no correlation between species composition at a site and siren presence/absence (RELATE: ρ = -0.211, p = 0.99) or siren abundance (ρ= -0.126, p = 0.99). Regardless of siren presence/absence or abundance, species composition between regions was different (ANOSIM: R = 0.478, p = 0.001).

Environmental Variables Water temperature, pH, conductivity, water depth, and dissolved oxygen measurements were collected at thirty-five of the thirty-six sampled water bodies between May, June, July and October 2014. Sites sampled in 2013 did not contain a complete dataset for environmental parameters for analysis. Water depth and dissolved oxygen readings were extremely variable within water bodies and time of day, and were not statistically analyzed. The mean water depth and standard error (SE) across sites was

26.64 ± 1.1 cm, with depths ranging from 10 to 98 cm. The mean dissolved oxygen and

SE was 8.98 ± 0.3 mg/L, and ranged from 0.50 to 17.80 mg/L. The pH values were rounded to the nearest 0.5 value. The mean pH and SE across sites was 8.4 ± 0.15, and ranged from 6.0 to 9.5. The mean conductivity and SE across sites was 1096 ± 92 μs/cm, and ranged from 99 to 1887 μs/cm. The mean water temperature and SE across sites was

31 ± 0.4 °C, and ranged from 23.6°C to 36°C.

The water temperature, pH, and conductivity were statistically analyzed across sites in an attempt to correlate siren presence/absence and the environmental parameters.

The BEST analysis revealed that no correlation between pH, conductivity, and water temperature with siren presence/absence (BEST: ρ = -0.029, p = 0.96) or abundance (ρ =

-0.049, p = 0.98) could be established.

Vegetation Composition No significant differences were found between percent cover (across all three vegetation zones as a whole) of sites with siren presence/absence (R = -0.205, p = 0.97) or abundance (R = -0.135, p = 0.99). There was a significant difference in percent cover between regions across all vegetation zones (R = 0.262, p = 0.007), but no significant

difference between percent cover and siren presence/absence (R = -0.115, p = 0.87,

Figure 13). Despite a lack of correlation, the SIMPER analysis revealed that sites with sirens were 73% similar in percent cover composition (all zones included) and that sirens preferred a high percent cover along the edge zone (≤ 95.45%). Although there was no significant correlation, sites with sirens present clustered together with regards to percent cover of Edge zone vegetation (Figure 14). The SIMPER for the vegetation groups revealed that thorny brush (23.96%), overhanging trees (19%), and short (< 1 m) grasses

(18.51%) were the primary contributors to the edge zone site similarity with siren presence.

There were no significant differences between sites with siren presence/absence and vegetation group presence/absence within a single zone (Aquatic: R = -0.018, p =

0.57; Emergent: R = -0.137, p = 0.98; Edge: R = -0.135, p = 0.97). A significant difference between zones was revealed based on the presence/absence of vegetation groups (R = 0.397, p = 0.001), but no significant difference between sites with siren presence/absence based on vegetation groups within all zones (R = -0.205, p = 0.97).

Phylogenetic Data mtDNA content The complete mitochondrial genomes (CMG) of 9 sirens were sequenced in this study. The Florida siren was sequenced, in addition to 8 Texas sirens. The 8 Texas sirens were chosen for the CMG sequencing by collection region and maximum genetic divergence as observed in single sequence fragments. The CMG sequence of the Florida

Siren consists of 16543 base pairs (bp). The 8 sirens from Texas had considerably variable CMG sequence lengths, ranging from 16537 to 17144 bp in length (Table 2). As

with most vertebrates, the Siren mtDNA genomes were AT rich (Table 3). The tRNA and gene order for all Siren CMGS in this study were identical to other salamander mitochondrial genomes and most amphibians, and gene rearrangement was not observed

(anuran CMGS may lack genes or have duplicate genes) (Zhang et al., 2005). All genomes contained 22 tRNAs, two rRNA subunits, and 13 protein coding genes. The

Siren CMGS also contained one intergenic spacer, one non-coding control region (D-

Loop), and an origin of light strand replication. All regions were encoded on the heavy strand (leading strand), except for eight tRNAs and the ND6 gene, which were encoded on the light strand (lagging strand) (Samuels et al., 2005).

CMGS phylogeny and sequence divergence Maximum Likelihood (ML) and Bayesian analyses yielded nearly identical tree topologies for all data sets, except for Cyt-b and IGS. Phylogenetic trees from the siren+outgroup data set strongly supported the conventional amphibian groupings for salamander families, and the caecilian order (Figure 15). As expected, Typhlonectes natans was the most genetically dissimilar from the salamander sequences (42.62% –

44.56%), based on the Kimura 2-parameter corrected genetic distance calculations. The sister-taxon relationship between salamander families Ambystomatidae (Ambystoma tigrinum tigrinum) and Dicamptodontidae (Dicamptodon atterimus) was strongly supported by both ML and Bayesian analyses (100 bootstrap (BS), 1.0 posterior probability (PP)), with less sequence divergence between one another than with any other sequence analyzed (26%, compared to ≥ 28%, Figure 15). All sequences from family

Sirenidae formed a strongly supported, monophyletic clade (100 BS, 1.0 PP), meaning the clade includes a common ancestor and all descendants of that ancestor. In addition,

this clade represented a sister-taxon to Ambystomatidae and Dicamptodontidae.

Pseudobranchus axanthus was most closely related to Siren intermedia and Siren lacertina within the siren family, but was still 20% and 21% divergent, respectively.

Surprisingly, S. intermedia and S. lacertina were more similar to one another, with a divergence of 10%, than they were to any other sirens in the dataset. This dataset revealed that sirens from Texas formed a well-supported, monophyletic clade (100 BS,

1.0 PP) distinct from previously identified siren species, S. intermedia and S. lacertina

(Figure 15). The mean genetic divergence between S. intermedia and all Texas sirens was

12% (11.6%-12.2%) and the mean genetic divergence between S. lacertina and all Texas sirens was 11.99% (11.5%-12.2%); however, divergence was found between three siren individuals referred to as sirens SS20, SS23, and ATT-1) within the Texas group, and between one another. Sirens SS20, SS23, and ATT-1 differed from the rest of the Texas sirens by 3.0%, 3.9%, and 2.8%, respectively, and on average they differed from one another by 3.2%.

The siren-only data set allowed us to examine the 9 siren CMGS (sequenced in this study) more extensively, due to the inclusion of intergenic spacer and flanking tRNAs (tRNATHR and tRNAPRO) and the D-Loop, which were not present in the siren-outgroup data set. The known S. lacertina from Florida sequenced in this study provided a good reference for interspecific divergence patterns because it also contains these gene regions. The ML and Bayesian trees for the siren-only dataset were similar, but not identical in topology (Figure 16). Both the Bayesian and ML trees showed the eight

Texas sirens forming a monophyletic group, with a distinct and strongly supported separation of sirens SS20, SS23, and ATT-1 (Figure 16). However, the 5 remaining

Texas sirens differed slightly in their terminal branch locations, but were still strongly supported as a phylogenetic entity (100 BS, 1.0 PP) (100 BS, 1.0 PP) (Figure 16).

In the siren-only analysis, the S. lacertina was highly divergent from the 8 sirens from Texas, with a mean genetic divergence of 12%, though divergence values ranged from 11.5% - 12.2%. These values were concordant with the interspecific divergence observed between S. lacertina and S. intermedia in the siren+outgroup dataset. Within the eight Texas sirens, distinct groupings and genetic separation were evident with sirens

SS20, SS23, and ATT-1 separating independently from the remaining 5 Texas sirens

(S24, SS6, SS27, SS29, SS3). These five Texas sirens showed little sequence variation across the entire mitochondrial genome, with a mean genetic divergence between of

0.6%. In contrast, sirens SS20, SS23 and ATT-1 separated independently from the remaining Texas sirens, but also formed independent and strongly supported branches from one another (Figure 16). The genetic divergence between both sirens SS23 and

SS20, and sirens SS23 and ATT-1 was 3.7%. The genetic divergence between sirens

ATT-1 and SS20 was slightly smaller, at 2.8%. The mean genetic divergence between

SS20, SS23, and ATT-1 with the remaining 5 Texas sirens was 3.5%, 4.1%, and 3.3%, respectively. The divergence values increased from the siren+outgroup analysis to the siren-only analysis by 0.4%, 0.2%, and 0.5%, for the relationships between sirens SS20,

SS23, ATT-1 and the Texas sirens, respectively. The divide between sirens SS20, SS23, and ATT-1 from the remaining Texas sirens increased with the inclusion of the D-Loop and the non-coding intergenic spacer.

Variability was observed in the length of the intergenic spacer (IGS) for all sirens.

In the siren-only analysis, the mtDNA IGS sequences from sirens SS20, SS23, and ATT-

1 had a significantly shorter non-coding intergenic spacer region than those from the remaining Texas sirens sequenced for this region. Sirens SS20 and SS23 both contained a non-coding region of 250 base pairs (bp). Siren ATT-1 contained a non-coding region of

447 bp, which is substantially smaller than the non-coding region for the remaining Texas sirens, ranging in length from 832-847 bp. The SG siren intergenic spacer was 250 bp, and the S. lacertina spacer was 251 base pairs long. The large intergenic spacer of the

Texas sirens mtDNA contained three repeat regions that varied in size from 200-203 bp, with single nucleotide changes and repeats. The repeat region of the Texas sirens begins at the same location, 63 bp from the end of tRNATHR with the repeated sequence element

5’- ATTTAGTC-3’. The sequences within the repeat regions were not documented anywhere else in the CMGS.

Gene phylogeny and sequence divergence Analysis of the 16S, CO1, Cyt-b, and IGS mtDNA gene regions of the Sirenidae family revealed incongruent and variable phylogenetic relationships. 16S mtDNA for P. axanthus, S. intermedia, S. lacertina and the unidentified voucher specimen (SG) from

Williamson County, Texas were included in the analysis, in addition to 11 Texas siren samples. The Bayesian and ML trees were topologically identical for the 16S region, rooted by P. axanthus, and strongly supported a monophyletic group of Texas sirens with

(99 BS, 1.0 PP). The trees also supported a monophyletic group for the entire Sirenidae family (Figure 17). S. intermedia and S. lacertina showed similar divergence patterns as the previous datasets, and were 7.7% genetically dissimilar. The SG voucher specimen was 7.9% divergent from S. intermedia, and was 4.3% divergent from S. lacertina. S. intermedia, S. lacertina and SG displayed mean genetic divergences from the Texas

sirens of 8.0%, 6.8%, and 3.7%, respectively. The topology did not show strong divergence between sirens SS20, SS23, and ATT-1, as previously indicated by the larger mitochondrial fragments; however, siren SS20 displayed a mean sequence divergence of

1.1% from the Texas sirens, compared to the 0.0%-0.8% range displayed between the remaining Texas sirens. Despite a 1.1% divergence, sirens SS20 and SS3, which were captured from the same water body, stemmed from a highly supported node (80 BS, 0.99

PP).

The CO1 sequence dataset most closely resembled the more encompassing mitochondrial genome dataset (Figure 18). The Bayesian tree and the ML tree were extremely similar, but differed in a few terminal locations of the Texas sirens (Figure 18).

For CO1, S. intermedia and S. lacertina stemmed from the same branch with moderate branch support (50 BS, 0.83 PP), but were genetically divergent by 10.9%. On average, the Texas sirens were 12.8% divergent from S. intermedia, and 13.6% divergent from S. lacertina. The ten Texas sirens analyzed created a monophyletic group with strong branch support (100 BS, 1.0 PP). Divergence within the sirens from Texas ranged from

0.0% to 1.6%. Again, sirens SS20 and SS3 grouped together, with a Bayesian branch support of 0.78 PP, and the remaining Texas sirens appeared to be more genetically similar to sirens from geographically close water bodies.

The Bayesian and ML tree topologies for Cyt-b differed slightly in the branching of S. intermedia and S. lacertina, yet both presented an intriguing structure. The Bayesian

Cyt-b analysis strongly supported separate branches (0.82-1.0 PP) for P. axanthus, S. intermedia, S. lacertina, and sirens SG, SS23, SS20, and ATT-1, but all in a monophyletic group (Figure 19). Both the Bayesian and ML tree strongly supported the

grouping of the remaining 13 sirens from Texas (99 BS, 1.0 PP). Siren intermedia and S. lacertina were 9.7% divergent within Cyt-b. Siren SS20 was 13.0% divergent from S. intermedia, 10.8% divergent from S. lacertina, 7.9% divergent from siren SS23, 5.9% divergent from siren ATT-1, and on average 10.6% (9.8-11.0%) divergent from all remaining Texas sirens. Siren SS23 was 15.2% divergent from S. intermedia, 12.6% divergent from S. lacertina, 10.3% divergent from siren ATT-1 and on average 12.8% divergent from the remaining Texas sirens. Siren ATT-1 was 14.0% divergent from S. intermedia, 12.3% divergent from S. lacertina, and on average 8.8% divergent from the remaining Texas sirens. Sequence divergence was much smaller within the 13 Texas sirens with a mean genetic variation of 0.8%, ranging from 0.0%-1.6%. The mean genetic variation between the 13 Texas sirens and S. intermedia was 15.9%, while with S. lacertina it was 14.3%.

The Bayesian tree and ML tree for the IGS region were identical in topology

(Figure 20). Both trees showed a clear divergence between sirens SS20, SS23, ATT-1 and SG from the remaining Texas sirens. The 4 divergent sirens were on a strongly supported branch (89 BS, 0.99 PP), as was the remaining group of Texas sirens (97 BS,

1.0 PP). The S. intermedia and P. axanthus CMG sequenes did not contain a sequenced

IGS region, and for this reason, the S. lacertina mtDNA (CMG) sequenced in this study served as the sole interspecific reference. The mean genetic distance between S. lacertina and the major Texas siren group was 23.4%, whereas, the mean genetic distance between

S. lacertina and sirens SS20, SS23, ATT-1, and SG was 30.4%. Within the divergent group of sirens, divergence values were high, except for sirens SS23 and SG, which were on the same node (1.0 PP) and showed a 0% genetic divergence. Within the divergent

group, siren ATT-1 was 17.7% divergent from sirens SS23 and SG, and 14.0% divergent from siren SS20. Siren SS20 was 10.0% divergent from sirens SS23 and SG. The mean genetic divergence between the divergent group and the Texas siren group was 26.7%.

Within the large Texas siren group the mean genetic divergence was 1.1%.

Isolation by Distance The isolation by distance analyses revealed a significant, albeit small, correlation between the genetic and geographical distances of the CO1 gene for the Texas sirens. For

CO1, the untransformed genetic distance (p-distance) and untransformed geographic distance matrix was significant (Mantel test P = 0.022, P < 0.05); however, the coefficient of determination from the regression analysis was low (r2 = 0.28, Figure 21), but comparable to other studies (MartíNez-Solano et al., 2007). The isolation by distance analyses also revealed a significant correlation for all analyses of Cyt-b for all Texas sirens. For Cyt-b, the untransformed genetic distance and untransformed geographic distance matrix (P = 0.032, r2 = 0.3), and the transformed genetic distance and untransformed geographic distance matrix (P = 0.01, r2 = 0.32) yielded significant, and relatively strong correlations. Analyses for 16S using all Texas sirens did not yield significant or correlated results, meaning siren genetic variation was not correlated with geographic location. For IGS, nearly all analyses were significant (P = 0.01 – 0.06), but only the untransformed genetic distance and untransformed geographic distance matrix

(r2 = 0.23) and the transformed genetic distance and untransformed geographic distance matrix (r2 = 0.22) showed any correlation. For the 16S, CO1, Cyt-b and IGS analyses, the siren sample sizes were 12, 10, 17, and 17 respectively.

Based on the slight genetic divergence of sirens SS20, SS23, ATT-1, and SG as described with the Kimura-corrected genetic distances, isolation by distance was also tested for each gene with the removal of these four sirens from analysis. The removal yielded sample sizes of 8 sirens for 16S, 7 sirens for CO1, 13 sirens for Cyt-b, and 13 sirens for IGS. The removal of sirens SS20, SS23, and ATT-1 from analysis for CO1 showed a stronger correlation (r2 = 0.45) between the transformed genetic distance with the transformed geographic distance, but the Mantel test proved to be non-significant (P

= 0.055). The isolation by distance without sirens SS20, SS23, ATT-1, and SG for Cyt-b revealed a significant correlation (P = 0.022) between the untransformed genetic distance and untransformed geographic distance matrix; however, the coefficient of determination from the regression analysis was also low (r2 = 0.246). Again, analyses for 16S with the removal of sirens SS20, SS23, ATT-1, and SG did not yield significant or correlated results. For the IGS region, with the removal of SS20, SS23, ATT-1, and SG, all four analyses were significant (P = 0.008 – 0.02), and slightly correlated (r2 = 0.107 – 0.169).

IV. DISCUSSION

Species Identification Goin (1957) proposed that the sirens in South Texas differed from Siren intermedia and Siren lacertina in several respects, but his observations were based on phenotypic characteristics, primarily coloration, costal groove number, and tail shape.

These differentiable subtleties seemingly represented a distinct siren group in South

Texas. This group was termed the “Rio Grande Siren,” and was a proposed subspecies of

Siren intermedia (Siren intermedia texana), though this was never validated by genetic means (Goin, 1957). Prior to this study, molecular analysis had never been conducted to determine the identification of siren species in South Texas.

The genetic findings from this study revealed that the sirens in South Texas are not a subspecies of S.intermedia, but are a genetically distinct group from both S. intermedia and S. lacertina. Goin (1957) appears correct in differentiating the sirens from

South Texas, but taxonomically misplaced the “Rio Grande Siren.” The genetic divergence observed in both the complete mitochondrial genome (CMG) sequences and the single mitochondrial gene sequences of the South Texas sirens sequenced in this study support at least one distinct species of siren in South Texas. The degree of genetic divergence observed between the South Texas sirens sequenced in this study and the S. intermedia and S. lacertina sequences, for the CMG and most single genes, is comparable to other studies that have differentiated and declared a distinct species (Funk et al., 2012;

Vences et al., 2005b; Vredenburg et al., 2007). This study also benefited from the use of photo-documentation and the Wild-ID software, which confirmed zero siren re-captures,

meaning repeat sampling did not contribute to the sequence similarities observed for the

South Texas sirens.

Previous studies have compared the utility of the CMG sequences against the utility of single genes for phylogenetic inference (Mueller et al., 2004). In this study, both

CMG sequences and single genes were analyzed independently to assess their utility in species identification. This study demonstrated that the CMG is a much more robust dataset than single genes for species identification, but that single genes may be useful for studying genetic relationships among populations (Prasad et al., 2008). The single genes analyzed in this study (16S, CO1, Cyt-b, and IGS) displayed varying degrees of sequence divergence between sirens, varying phylogenetic tree topologies, and varying placement for the individual Texas sirens SS20, SS23, and ATT-1, likely due to differing mutation and evolutionary rates within and between genes (Gibson, 2004; Rubinoff et al.,

2006).

The Cyt-b and the intergenic spacer (IGS) regions proved insufficient in siren species delimitation and in overall siren phylogeny in this study. Cyt-b and the IGS have most commonly been used for phylogenetic resolution, but are known to display extreme variability and rapid evolution (Kuchta and Tan, 2004; MartíNez-Solano et al., 2007;

Matsui et al., 2007). Because the IGS is a non-coding region, substantial mutations may not affect siren function and could be useful for phylogenetic inference, but are too variable for accurate species identification (McKnight and Shaffer, 1997). Because single genes within the CMG experience varying degrees of nucleotide mutation based on evolutionary pressures, genes with low rates of molecular evolution more accurately display true phylogenetic relationships than rapidly evolving genes, suggesting this may

also be true for species identification (Mueller et al., 2004). This suggests that Cyt-b may not be an ideal tool for species identification because the large genetic divergences displayed may be misleading for speciation (Johns and Avise, 1998).

Despite being cautioned for use with amphibians, CO1 was the most useful single gene for species identification in this study. CO1 is known as the “universal barcode” for vertebrates, but has been debated as a useful marker for amphibians because of its proposed high nucleotide variability, specifically in the third codon position (Mueller et al., 2004; Ratnasingham and Hebert, 2007; Vences et al., 2005a). CO1 has been considered a misleading gene because of its contribution to an explosion of “new” amphibian species (Smith et al., 2008). Instead, 16S has been proposed as a more suitable marker for amphibian studies due to its preferential priming sites, and ability to discern between closely related species (Vences et al., 2005a; Vences et al., 2005b). Studies including frogs and some salamanders, have displayed vast intraspecific divergences within CO1, but this study demonstrated that the CO1 gene exhibited extremely low intraspecific genetic divergence (0% - 1.6%) between all Texas sirens sampled for CO1, and a distinctly greater interspecific divergence between the Texas sirens in this study and S. intermedia and S. lacertina (≥ 12.8%) (Grosjean et al., 2015; McKnight and

Shaffer, 1997; Mueller et al., 2004; Vences et al., 2005a). Similar to Funk’s findings

(2015), both the CO1 and 16S datasets revealed relatively concordant tree topologies and similar genetic divergences, which in this study, closely mirrored the CMG dataset and presented a monophyletic group for the South Texas sirens. CO1 also seems to be useful for discerning between genetic divergences based on geographic distance (IBD) as specifically as site location, as displayed in Figure 18. Both sirens SS20 and SS3 were

collected from the same region, and despite siren SS20’s overall CMG sequence divergence, both sirens group together (Figure 22). These results suggest that if the CMG cannot be sequenced for species identification, CO1 is a useful single gene for conservative siren species identification.

One interesting finding from this study relates to the unidentified voucher specimen (SG) collected from the San Gabriel River in central Texas. Anecdotal reports have documented the presence of fish tank stones along the bank of the San Gabriel river where the siren was collected. This initially led us to hypothesize that the voucher specimen may have been a released pet, but the genetic analysis reveals perplexing data.

For SG, the 12S, CYTB, ND5, and ND6 genes show a closer phylogenetic relationship to

S. lacertina, but the intergenic spacer shows a closer phylogenetic relationship to the

Texas siren SS23 (0%), and 16S reveals genetic similarities between both S. lacertina

(4.3%) and all other sequenced Texas sirens (3.7%). This suggests that SG may simply be an extremely divergent individual from the majority of the South Texas sirens sampled, that SG may represent a candidate species of siren in Texas, or that SG may be more closely related to the sirens in East Texas (proposed as Siren intermedia nettingi); however, a much more robust sample size is necessary for validation, and as discerned from this study, species identification may only be determined with the CMGS.

Distribution The distribution of sirens throughout Texas has not been well defined. Goin

(1957) suggested that the “Rio Grande Siren” primarily occupied the Rio Grande River

Basin and that the proposed S. intermedia nettingi inhabited East Texas, but speculated

that there was a questionable region of species overlap near Kingsville, TX. From this study, we know that the genetic divergence displayed between sirens does not stem from geographic isolation, and that there is at least one species of siren present in South Texas, with a distribution extending from the Lower Rio Grande Valley of Texas to at least

Kleberg county, as Goin (1957) suggested. Because intraspecific divergence can be high for amphibians, it is highly likely that all of the sirens in this study from Texas are from the same species; however, the genetic divergence of sirens SS20, SS23, ATT-1 and SG requires more analysis. With more sampling, from a more extensive range, greater resolution may be available to define the specie(s) of sirens in South Texas and to determine species boundaries.

Habitat Characteristics Species composition, environmental variables, and vegetation composition did not differ significantly between sites with siren presence/absence, as was expected, but seasonality and percent edge cover were important factors for siren presence in South

Texas. Co-occurring species were predicted to correlate to siren presence based on siren diet and potentially from predatory prey interactions (Davic and Jr., 2004). In addition, environmental variables were predicted to correlate with siren presence, based on studies that have found correlations between siren capture abundance with water depth and water temperature (Schalk and Luhring, 2010; Sorensen, 2004). Terrestrial salamander studies have often shown that vegetation affects siren distribution, as have siren specific studies, with preference for vegetation type and location within the water body (Davic and Jr.,

2004; Schalk et al., 2010). The uncorrelated results from this study could be a result of

small sampling size (36 sites) or water body connectivity, but were seemingly produced from a very low siren capture abundance based on the extreme seasonality in South

Texas.

Sirens were collected in water bodies with a high percent of edge vegetation cover

(≥95%). Because habitat loss and fragmentation are among the leading causes for amphibian population declines, this information is extremely useful for management implications in South Texas (Stuart, 2004). South Texas is undergoing a number of restoration efforts for city-managed “resacas,” which include dredging, bulk-heading, and non-invasive vegetation removal, which could all potentially impact percent cover of edge vegetation (Aldridge, 2000). Bulk-heads are constructed to prevent erosion and land loss, which can be caused by large rainfalls or invasive species, such as nutria and armored catfish (Nico and Martin, 2001; Wildlife Services, 2010). Bulk-heading removes the gradual slope of the water body edge and subsequently the edge vegetation.

In addition, invasive grasses, such as guinea grass, are being removed from water body edges in South Texas with herbicides or by mechanical means (Van Devender et al.,

2006). This vegetation removal could negatively impact sirens because 18.5% of the edge vegetation consisted of grasses (Van Devender et al., 2006). The high percent of edge vegetation from sites with siren captures suggests that bulk-heading and vegetation removal could negatively impact siren populations because of the lack of vegetation cover, and should be assessed for land management practices.

Seasonal sampling from this study yielded drastic differences in siren abundance and catch-per-unit-effort (CPUE), and produced results that differed from historical siren capture records in Texas (Figure 5, Figure 6, Figure S3). Historical records of siren

captures in South Texas were minimal from May to November and the lack of CPUE data made it unclear whether these low yields were due to a lack of sampling effort, or whether the low yields were results based on seasonality (Figure S3). Based on the results from this study, it appears that siren capture rates and abundance differed in South Texas because of season, but more sampling is necessary to confirm this.

Results from this study demonstrated that siren abundance and CPUE rates were greater in the fall (September and October) than in the summer (June, July, and August).

Seasonal variation is not uncommon for amphibians, and abundances often peak after substantial rainfall (Duellman, 1995). This information supports the historical records

(1950 - present) of siren captures in South Texas (Starr, Hidalgo, Cameron, Willacy,

Kleberg, and Jim Wells Counties) for the summer, but strongly contradicts the historical records for the fall (Figure S3). Additionally, this study differs from other siren studies, such as Sorensen’s (2004) Florida study, which collected the greatest abundance of sirens in July, with low yields in October. Based on the historical records for South Texas in addition to the results from this study, sirens in South Texas exhibit extreme seasonality

(Figure S3).

The seasonal variation and low capture abundances provided knowledge about sirens in South Texas that was previously unknown. Because the few sirens were collected during the summer time and all were relatively small, it is likely that larger sirens may have begun to aestivate. Sirens are known to be capable of aestivating, but the exact triggers for aestivation are not known. Perhaps the decreased water levels and the rise in water temperatures during summer triggered the large sirens to burrow into the sediment, explaining the low catch abundance. In contrast to large sirens, smaller sirens

have less body fat, and would not be able to survive for an extended period of time in a dormant state, thus delaying their ability to aestivate (Gehlbach et al., 1973).

From this study, the data suggests that sampling for sirens during the summer time in South Texas will not yield an accurate account of siren presence or overall population abundance, which could ultimately affect conservation plans. This was made extremely evident by sampling the same water body for a consecutive period of time, and capturing a small abundance, despite knowing sirens were present. In contrast, the fall displayed a more accurate representation of siren site presence, promoting the fall as an ideal season for siren population assessment in South Texas. Thus, future population assessments for conservation and management plans should be conducted in the fall, and may benefit from the use of eDNA (environmental DNA), which could assist with determining siren presence in water bodies (Dejean et al., 2012).

V. CONCLUSION

This study provides the first documented attempt to study siren habitat in South

Texas, and to assess species identification of sirens in South Texas utilizing molecular techniques. Between 2013 and 2015, siren collection throughout South Texas enabled the identification of at least one distinct, and has contributed to the sparse knowledge surrounding siren inhabitance in South Texas.

Understanding the characteristics of a preferred habitat of amphibians is critical as populations continue to decline (Beebee and Griffiths, 2005; Cushman, 2006); however, for this study, co-occurring species, chemical parameters, and vegetation composition were not primary contributors to siren habitat selection (Dodd and Smith, 2003; Petranka,

2010; Scheele et al., 2014; Snodgrass et al., 1999). Seasonality and percent edge cover findings from this study will provide extremely useful baseline data for future research, and will aid in assessing siren population abundance in South Texas. For future molecular analysis, the inclusion of other tools, such as nuclear genes and morphological descriptions, may present an even more comprehensive understanding of the underlying evolutionary processes contributing to siren species identification.

The results from this study have contributed significantly to the understanding of sirens in South Texas, and should be utilized to continue population assessments, further explore species identification, define species distribution, and ultimately develop land management and conservation plans.

VI. LITERATURE CITED

Aldridge, D. C. (2000). The impacts of dredging and weed cutting on a population of freshwater mussels. Biological Conservation 95, 247–257.

Arif, I. A. and Khan, H. A. (2009). Molecular markers for biodiversity analysis of wildlife animals: a brief review. Animal Biodiversity and Conservation 32, 9–17.

Beebee, T. J. C. and Griffiths, R. A. (2005). The amphibian decline crisis: A watershed for conservation biology? Biological Conservation 125, 271–285.

Bendik, N. F., Morrison, T. A., Gluesenkamp, A. G., Sanders, M. S. and O’Donnell, L. J. (2013). Computer-Assisted Photo Identification Outperforms Visible Implant Elastomers in an Endangered Salamander, Eurycea tonkawae. PLoS ONE 8, e59424.

Bohonak, A. J. (2002). IBD (Isolation By Distance): a program for analyses of isolation by distance. Journal of Heredity 153–154.

Bolger, D., Morrison, T., Vance, B., Lee, D. and Farid, H. (2012). A computer- assisted system for photographic mark–recapture analysis. Methods in Ecology and Evolution 3, 813–822.

Branicki, W., Kupiec, T. and Pawlowski, R. (2003). Validation of cytochrome b sequence analysis as a method of species identification. Journal of Forensic Sciences 48, 83–87.

Brown, B. C. (1950). An annotated check list of the reptiles and amphibians of Texas. 1st ed. Baylor University Press.

Chan, L. M. (2003). Seasonality, microhabitat and cryptic variation in tropical salamander reproductive cycles. Biological Journal of the Linnean Society 78, 489–496.

Chauncey Bishop, S. (1943). Handbook of Salamanders: The Salamanders of the United States, of Canada. Cornell University Press.

Collette, B. B. and Gehlbach, F. R. (1961). The Salamander Siren intermedia intermedia Leconte in North Carolina. Herpetologica Vol. 17, 203–204.

Conant, R. and Collins, J. T. (1998). A Field Guide to Reptiles & Amphibians: Eastern and Central North America. Third. Houghton Mifflin Company.

Cummings, M. P., Otto, S. P. and Wakeley, J. (1995). Sampling properties of DNA sequence data in phylogenetic analysis. Molecular Biology and Evolution 12, 814–822.

Cushman, S. A. (2006). Effects of habitat loss and fragmentation on amphibians: A review and prospectus. Biological Conservation 128, 231–240.

Davic, R. D. and Jr., H. H. W. (2004). On the Ecological Roles of Salamanders. Annual Review of Ecology, Evolution, and Systematics 35, 405–434.

Dejean, T., Valentini, A., Miquel, C., Taberlet, P., Bellemain, E. and Miaud, C. (2012). Improved detection of an alien invasive species through environmental DNA barcoding: the example of the American bullfrog Lithobates catesbeianus: Alien invasive species detection using eDNA. Journal of Applied Ecology 49, 953–959.

Delany, M. F. and Abercrombie, C. L. (1986). American Alligator Food Habits in Northcentral Florida. The Journal of Wildlife Management 348–353.

Dodd, C. K. J. and Smith, L. L. (2003). Habitat destruction and alteration. In Amphibian Conservation, p. 336. Smithsonian Institution Press.

Duellman, W. E. (1995). Temporal Fluctuations in Abundances of Anuran Amphibians in a Seasonal Amazonian Rainforest. Journal of Herpetology 29, 13.

Duke, J. T. and Ultsch, G. R. (1990). Metabolic oxygen regulation and conformity during submergence in the salamanders Siren lacertina, Amphiuma means, and Amphiuma tridactylum, and a comparison with other giant salamanders. Oecologia Vol. 84, 16–23.

Edgar, R. C. (2004). MUSCLE: multiple sequence alignment with high accuracy and high throughput. Nucleic Acids Research 32, 1792–1797.

Etheridge, K. (1990). Water Balance in Estivating Sirenid Salamanders (Siren lacertina). Herpetologica Vol. 46, pp. 400–406.

Fauth, J. E. and Resetarits Jr., W. J. (1999). Biting in the Salamander Siren intermedia intermedia: Courtship Component or Agonistic Behavior? Journal of Herpetology Vol. 33, pp. 493–496.

Fishman, A. P., Galante, R. J., Winokur, A. and Pack, A. I. (1992). Estivation in the African Lungfish. Proceedings of the American Philosophical Society 136, 61–72.

Fouquet, A., Gilles, A., Vences, M., Marty, C., Blanc, M. and Gemmell, N. J. (2007). Underestimation of Species Richness in Neotropical Frogs Revealed by mtDNA Analyses. PLoS ONE 2, e1109.

Funk, W. C., Caminer, M. and Ron, S. R. (2012). High levels of cryptic species diversity uncovered in Amazonian frogs. Proceedings of the Royal Society B: Biological Sciences 279, 1806–1814.

Gehlbach, F. R. and Kennedy, S. E. (1978). Population Ecology of a Highly Productive Aquatic Salamander (Siren intermedia). The Southwestern Naturalist 23, 423– 429.

Gehlbach, F. R., Gordon, R. and Jordan, J. B. (1973). Aestivation of the salamander, Siren intermedia. American Midland Naturalist 455–463.

Gibson, A. (2004). A Comprehensive Analysis of Mammalian Mitochondrial Genome Base Composition and Improved Phylogenetic Methods. Molecular Biology and Evolution 22, 251–264.

Godley, J. S. (1983). Observations on the Courtship, Nests and Young of Siren intermedia in Southern Florida. American Midland Naturalist Vol. 110, pp. 215– 219.

Goin, C. J. (1957). Description of a new salamander of the genus Siren from the Rio Grande. Herpetologica 37–42.

Grosjean, S., Ohler, A., Chuaynkern, Y., Cruaud, C. and Hassanin, A. (2015). Improving biodiversity assessment of anuran amphibians using DNA barcoding of tadpoles. Case studies from Southeast Asia. Comptes Rendus Biologies 338, 351–361.

Guha, S., Goyal, S. P. and Kashyap, V. K. (2006). Genomic variation in the mitochondrially encoded cytochrome b (MT-CYB) and 16S rRNA (MT-RNR2) genes: characterization of eight endangered Pecoran species. Animal Genetics 37, 262–265.

Hampton, P. M. (2009). Ecology of the Lesser Siren, Siren intermedia, in an Isolated Eastern Texas Pond. Journal of Herpetology Vol. 43, pp. 704–709.

Hanlin, H. G. (1978). Food Habits of the Greater Siren, Siren lacertina, in an Alabama Coastal Plain Pond. Copeia, American Society of Ichthyologists and Herpetologists (ASIH) Vol. 1978, pp. 358–360.

Helm-Bychowski, K. M., Higuchi, R. G., Palumbi, S. R., Prager, E. M., Sage, R. D. and Stoneking, M. (1985). Mitochondrial DNA and two perspectives on evolutionary genetics. Biol J Linn Soc 26, 375–400.

Hill, R. L., Mendelson, J. R. and Stabile, J. L. (2015). Direct Observation and Review of Herbivory in Sirenidae (Amphibia: Caudata). Southeastern Naturalist 14, N5– N9.

James P. Collins and Andrew Storfer (2003). Global amphibian declines: sorting the hypotheses. Diversity and Distributions 9, 89–98.

Jensen, J. L., Bohonak, A. J. and Kelley, S. T. (2005). Isolation by distance, web service. BMC genetics 6, 13.

Jiang, L., Wang, G., Tan, S., Gong, S., Yang, M., Peng, Q., Peng, R. and Zou, F. (2013). The complete mitochondrial genome sequence analysis of Tibetan argali (Ovis ammon hodgsoni): Implications of Tibetan argali and Gansu argali as the same subspecies. Gene 521, 24–31.

Johns, G. C. and Avise, J. C. (1998). A comparative summary of genetic distances in the vertebrates from the mitochondrial cytochrome b gene. Molecular Biology and Evolution 15, 1481–1490.

Kimura, M. (1980). A simple method for estimating evolutionary rates of base substitutions through comparative studies of nucleotide sequences. Journal of molecular evolution 16, 111–120.

Köhler, J., Vieites, D. R., Bonett, R. M., García, F. H., Glaw, F., Steinke, D. and Vences, M. (2005). New Amphibians and Global Conservation: A Boost in Species Discoveries in a Highly Endangered Vertebrate Group. BioScience 55, 693–696.

Kuchta, S. R. and Tan, A.-M. (2004). Isolation by distance and post-glacial range expansion in the rough-skinned newt, Taricha granulosa: Phylogeography of the rough-skinned newt. Molecular Ecology 14, 225–244.

Lamb, T., Sullivan, B. K. and Malmos, K. (2000). Mitochondrial gene markers for the hybridizing toads Bufo microscaphus and Bufo woodhousii in Arizona. Journal Information 2000,.

Lanfear, R., Calcott, B., Kainer, D., Mayer, C. and Stamatakis, A. (2014). Selecting optimal partitioning schemes for phylogenomic datasets. BMC Evolutionary Biology 14, 82.

Lopez, C. H. and Brodie, E. D. (1977). The function of costal grooves in salamanders (Amphibia, Urodela). Journal of Herpetology 372–374.

Luhring, T. M. (2008). Population ecology of greater siren, Siren lacertina.

Lunt, D. H., Whipple, L. E. and Hyman, B. C. (1998). Mitochondrial DNA variable number tandem repeats (VNTRs): utility and problems in molecular ecology. Molecular Ecology 7, 1441–1455.

Maps, I. V. pGEMŪ-T and pGEMŪ-T Easy Vector Systems.

MartíNez-Solano, I., Jockusch, E. L. and Wake, D. B. (2007). Extreme population subdivision throughout a continuous range: phylogeography of Batrachoseps

attenuatus (Caudata: Plethodontidae) in western North America. Molecular Ecology 16, 4335–4355.

Matsui, M., Tominaga, A., Hayashi, T., Misawa, Y. and Tanabe, S. (2007). Phylogenetic relationships and phylogeography of Hynobius tokyoensis (Amphibia: Caudata) using complete sequences of cytochrome b and control region genes of mitochondrial DNA. Molecular Phylogenetics and Evolution 44, 204–216.

McKnight, M. L. and Shaffer, H. B. (1997). Large, rapidly evolving intergenic spacers in the mitochondrial DNA of the salamander family Ambystomatidae (Amphibia: Caudata). Molecular Biology and Evolution 14, 1167–1176.

Morrison, T. A. and Bolger, D. T. (2012). Wet season range fidelity in a tropical migratory ungulate. Journal of Animal Ecology 81, 543–552.

Mueller, R. L., Macey, J. R., Jaekel, M., Wake, D. B. and Boore, J. L. (2004). Morphological homoplasy, life history evolution, and historical biogeography of plethodontid salamanders inferred from complete mitochondrial genomes. Proceedings of the National Academy of Sciences of the United States of America 101, 13820–13825.

NatureServe (2013). NatureServe Explorer: An online encyclopedia of life.

Nico, L. G. and Martin, R. T. (2001). The South American Suckermouth Armored Catfish, Pterygoplichthys anisitsi (Pisces: Loricaridae), in Texas, with Comments on Foreign Fish Introductions in the American Southwest. The Southwestern Naturalist 46, 98.

Oscar Flores Villela and Ronald A. Brandon (1992). Siren lacertina (Amphibia, Caudata) in northeastern Mexico and southern Texas. Annals of Carnegie Museum Vol. 61, Pp. 289–291.

Parra-Olea, G., Wake, D. and Hammerson, G. A. (2008). Siren lacertina.

Pechmann, J. H. K. and Wilbur, H. M. (1994). Putting Declining Amphibian Populations in Perspective: Natural Fluctuations and Human Impacts. Herpetologica 50, 65–84.

Peterson, M. S. and VanderKooy, S. J. (1997). Distribution, Habitat Characterization, and Aspects of Reproduction of a Peripheral Population of Bluespotted Sunfish Enneacanthus gloriosus (Holbrook).pdf. Journal of Freshwater Ecology 12, 151– 162.

Petranka, J. W. (2010). Salamanders of the United States and Canada. Smithsonian Books.

Prasad, A. B., Allard, M. W., NISC Comparative Sequencing Program and Green, E. D. (2008). Confirming the Phylogeny of Mammals by Use of Large Comparative Sequence Data Sets. Molecular Biology and Evolution 25, 1795– 1808.

Ratnasingham, S. and Hebert, P. D. (2007). BOLD: The Barcode of Life Data System (http://www. barcodinglife. org). Molecular ecology notes 7, 355–364.

Raymond, L. A. (1991). Seasonal Activity of Siren intermedia in Northwestern Louisiana (Amphibia: Sirenidae). The Southwestern Naturalist Vol. 36, pp. 144– 147.

Ronquist, F. and Huelsenbeck, J. P. (2003). MrBayes 3: Bayesian phylogenetic inference under mixed models. Bioinformatics 19, 1572–1574.

Rubinoff, D., Cameron, S. and Will, K. (2006). A Genomic Perspective on the Shortcomings of Mitochondrial DNA for “Barcoding” Identification. Journal of Heredity 97, 581–594.

Rubinoff, D., San Jose, M., Johnson, P., Wells, R., Osborne, K. and Le Roux, J. J. (2015). Ghosts of glaciers and the disjunct distribution of a threatened California moth (Euproserpinus euterpe). Biological Conservation 184, 278–289.

Samuels, A. K., Weisrock, D. W., Smith, J. J., France, K. J., Walker, J. A., Putta, S. and Voss, S. R. (2005). Transcriptional and phylogenetic analysis of five complete ambystomatid salamander mitochondrial genomes. Gene 349, 43–53.

Schalk, C. M. and Luhring, T. M. (2010). Vagility of Aquatic Salamanders: Implications for Wetland Connectivity. Journal of Herpetology Vol. 44, pp. 104– 109.

Schalk, C. M., Luhring, T. M. and Crawford, B. A. (2010). Summer microhabitat use of the Greater Siren (Siren lacertina) and Two-toed Amphiuma (Amphiuma means) in an isolated wetland. Amphibia-Reptilia 31, 251–256.

Scheele, B. C., Boyd, C. E., Fischer, J., Fletcher, A. W., Hanspach, J. and Hartel, T. (2014). Identifying core habitat before it’s too late: the case of Bombina variegata, an internationally endangered amphibian. Biodiversity and Conservation 23, 775– 780.

Sievers, F. and Higgins, D. G. (2014). Multiple Sequence Alignment Methods. Humana Press.

Smith, M. A., Poyarkov, N. A. and Hebert, P. D. N. (2008). DNA BARCODING: CO1 DNA barcoding amphibians: take the chance, meet the challenge. Molecular Ecology Resources 8, 235–246.

Snodgrass, J. W., Ackerman, J. W., Bryan Jr, A. L. and Burger, J. (1999). Influence of hydroperiod, isolation, and heterospecifics on the distribution of aquatic salamanders (Siren and Amphiuma) among depression wetlands. Copeia 107– 113.

Sorensen, K. (2004). Population Characteristics of Siren lacertina and Amphiuma means in North Florida. Southeastern Naturalist 3, 249–258.

Steinfartz, S., Kusters, D. and Tautz, D. (2004). Isolation and characterization of polymorphic tetranucleotide microsatellite loci in the Fire salamander Salamandra salamandra (Amphibia: Caudata). Molecular Ecology Notes 4, 626–628.

Stuart, S. N. (2004). Status and Trends of Amphibian Declines and Extinctions Worldwide. Science 306, 1783–1786.

Tamura, K., Stecher, G., Peterson, D., Filipski, A. and Kumar, S. (2013). MEGA6: Molecular Evolutionary Genetics Analysis Version 6.0. Molecular Biology and Evolution 30, 2725–2729.

The IUCN Red List of Threatened Species (2015).

Tipton, B. L., Hibbitts, T. L., Hibbitts, T. D., Hibbits, T. J. and Laduc, T. J. (2012). Texas Amphibians: A Field Guide. 1st ed. The University of Texas Press.

Ultsch, G. R. (1973). Observations on the Life History of Siren lacertina. Herpetologica Vol. 29, pp. 304–305.

Van Devender, T. R., Espinosa-García, F. J., Harper-Lore, B. L. and Hubbard, T. (2006). INVASIVE PLANTS ON THE MOVE: CONTROLLING THEM IN NORTH AMERICA.

Vences, M., Thomas, M., Van der Meijden, A., Chiari, Y. and Vieites, D. R. (2005a). Comparative performance of the 16S rRNA gene in DNA barcoding of amphibians. Frontiers in Zoology 2, 5.

Vences, M., Thomas, M., Bonett, R. M. and Vieites, D. R. (2005b). Deciphering amphibian diversity through DNA barcoding: chances and challenges. Philosophical Transactions of the Royal Society B: Biological Sciences 360, 1859–1868.

Vredenburg, V. T., Bingham, R., Knapp, R., Morgan, J. A. T., Moritz, C. and Wake, D. (2007). Concordant molecular and phenotypic data delineate new taxonomy and conservation priorities for the endangered mountain yellow-legged frog: Phylogeography and decline of Rana muscosa and Rana sierrae. Journal of Zoology 271, 361–374.

Weisrock, D., Harmon, L. and Larson, A. (2005). Resolving Deep Phylogenetic Relationships in Salamanders: Analyses of Mitochondrial and Nuclear Genomic Data. Systematic Biology 54, 758–777.

Werler, J. E. and Dixon, J. R. (2000). Texas snakes: identification, distribution, and natural history. University of Texas Press.

Wildlife Services (2010). Nutria, an Invasive Rodent.

Yu, L., Li, Y.-W., Ryder, O. A. and Zhang, Y.-P. (2007). Analysis of complete mitochondrial genome sequences increases phylogenetic resolution of bears (Ursidae), a mammalian family that experienced rapid speciation. BMC Evolutionary Biology 7, 198.

Zhang, P. and Wake, D. B. (2009). Higher-level salamander relationships and divergence dates inferred from complete mitochondrial genomes. Molecular Phylogenetics and Evolution 53, 492–508.

Zhang, P., Zhou, H., Liang, D., Liu, Y.-F., Chen, Y.-Q. and Qu, L.-H. (2005). The complete mitochondrial genome of a tree frog, Polypedates megacephalus (Amphibia: Anura: Rhacophoridae), and a novel gene organization in living amphibians. Gene 346, 133–143.

Zhang, P., Papenfuss, T. J., Wake, M. H., Qu, L. and Wake, D. B. (2008). Phylogeny and biogeography of the family Salamandridae (Amphibia: Caudata) inferred from complete mitochondrial genomes. Molecular Phylogenetics and Evolution 49, 586–597.

Fragment Primer Name Sequence 5’  3’ Source Name T7 TAATACGACTCACTATAGGG Universal primer SP6 ATTTAGGTGACACTATAG Universal primer

L1-C2 L1-FW-1 ACACCGCCCATCACCCTCA This study L2-RV GACCTGGATTACTCCGGTCTGAACTC (Zhang et al., 2008) A-FW GGTTTACGACCTCGATGTTGGATCA (Zhang et al., 2008) A-RV GGTATGGGCCCAARAGCTT (Zhang et al., 2008) B-FW-3 GGACTTGAACCTACCCTAAAGAG This study B-FW-INT2* CGGAAACCCACCACTATTCT This study B-RV AGGGTGCCRATRTCYTTRTGRTT (Zhang et al., 2008) C1-FW-2 GAAGCCCCGGCAGATTCTAT This study C1-FW-INT1* GGGCTACAACCCTGCACCTA This study C2-RV-2 CCCTGCTAACCCTAAGAAATGTTGTGG This study

C2-G C2-FW-1 TGTCTTGTCTATGGGGGCAGTATTTGC This study E-FW-NEW2 TGCCATAGACGCACAAGAAAT This study F-FW AAGCAATAGCCTTTTAAGC (Zhang et al., 2008) MIDDLE-F-1 AGGTCAGCATTCGCCAAGTT This study F-RV AACCRAAATTTAYTRAGTCGAAAT (Zhang et al., 2008) AFTER-F-1 CTCTGCCCGTCTTCCATTCT This study G-FW ATTTCGACTYAGTAAATTTYGGTT (Zhang et al., 2008) G-MONEY-FW CACTTATTATTATTACCCGATGAG This study G-RV-2INT GGCGTGTCATCAGCCAATTA This study

G-L1 K9FW-GTOL1 ACCCTCTTTACAAACCGAGAAGG This study NADH5-4-INT CGCCGAAGCTAACACCGCAGC This study NADH5-3-LAC* TTTAGCAGCAATGGGAAAATCAGC This study I-FW-INT1 ATCGCAACATCATTCACAGCAGTT This study I-FW* ATTGTAGCATTTTCAACATC (Zhang et al., 2008) I-FW-INT2* AATGATTGGGGAGGAGTTGGTGAA This study NOMATCH-1 GCACGCAAAGACCCTAATGA This study M1-FW GAAAAACCAAYGTTGTATTCAACTATAA (Zhang et al., 2008) ENDGAP-3-INT CCCATTGGTTACCCCACCTCAC This study ENDGAP-1* TTGCCTACGCCATTCTTCG This study ENDGAP-4-LAC* AACCTACTTGGAGACCCTGAA This study EXTRA-RV-2 CAGTAGCTGCCGTCTTGGTG This study EXTRA-RV-1* CATTTTGTGCCGACCCCTAT This study OTHERGAP-3-INT TGTGGCTGGTTAGTCCAAGAG This study OTHERGAP-1-S31* AATGTCACGCCGATAAGGAAG This study M2-RV TCGATTATAGAACAGGCTCCTCT (Zhang et al., 2008) K10RV-GTOL1 CATCCCACTCTTTTGCCACAG This study

Table 2. Annotation and gene organization of the complete mitochondrial genomes of 8 Siren spp. from Texas and one Siren lacertina from Florida sequenced in this study. An asterisk (*) denotes the complement sequencing strand (light strand). Siren spp. individuals Gene S24 SS27 SS29 SS6 SS3 SS20 SS23 ATT-1 SL4 tRNAPHE 1..68 1..68 1..68 1..68 1..68 1..68 1..68 1..68 1..68 12S 69..1005 69..1005 69..1005 69..1005 69..1005 69..1016 69..1005 69..1001 69..1005 tRNAVAL 1007..1076 1007..1076 1007..1076 1007..1076 1007..1076 1018..1087 1007..1076 1003..1072 1007..1076 16S 1077..2656 1077..2657 1077..2657 1077..2657 1077..2656 1088..2668 1077..2657 1073..2652 1077..2660 tRNALEU1 2657..2731 2658..2732 2658..2732 2658..2732 2657..2731 2669..2743 2658..2732 2653..2727 2661..2735 ND1 2732..3694 2733..3695 2733..3695 2733..3695 2732..3694 2744..3706 2733..3695 2728..3690 2736..3698 tRNAILE 3694..3763 3695..3764 3695..3764 3695..3764 3694..3763 3706..3775 3695..3764 3690..3759 3698..3767 tRNAGLN 3764..3833 3765..3831 3765..3833 3765..3833 3764..3832 3776..3845 3765..3832 3760..3829 3768..3837 tRNAMET 3833..3900 3831..3898 3833..3900 3833..3900 3832..3899 3845..3912 3832..3899 3829..3896 3837..3904 ND2 3902..4944 3900..4942 3902..4944 3902..4944 3901..4943 3914..4956 3901..4943 3898..4940 3906..4940 tRNATRP 4944..5013 4942..5011 4944..5013 4944..5013 4943..5012 4956..5025 4943..5012 4940..5009 4940..5002 tRNAALA 5013..5081* 5011..5079* 5013..5081* 5013..5081* 5012..5080* 5025..5093* 5012..5080* 5009..5077* 5002..5066* tRNAASN 5084..5155* 5082..5153* 5084..5155* 5084..5155* 5083..5154* 5096..5167* 5083..5154* 5080..5151* 5069..5140*

OL 5156..5191 5154..5189 5156..5191 5156..5191 5155..5190 5168..5203 5155..5191 5152..5187 5141..5177 tRNACYS 5192..5256* 5190..5254* 5192..5256* 5192..5256 5191..5255 5204..5268 5192..5256 5188..5252 5178..5242 tRNATYR 5257..5324* 5255..5320* 5257..5323* 5257..5324* 5256..5322* 5269..5336 5257..5324 5253..5320 5243..5310 COX1 5326..6874 5322..6870 5325..6873 5326..6874 5324..6872 5338..6886 5326..6874 5322..6870 5312..6860 tRNASER1 6875..6944* 6871..6940* 6874..6943* 6875..6944* 6873..6942* 6887..6956* 6875..6944* 6871..6940* 6861..6930* tRNAASP 6950..7017 6946..7013 6949..7016 6950..7017 6948..7015 6962..7029 6950..7017 6946..7013 6936..7003 COX2 7021..7708 7017..7704 7020..7707 7021..7708 7019..7706 7033..7720 7021..7708 7017..7704 7007..7694 tRNALYS 7709..7783 7705..7779 7708..7782 7709..7783 7707..7781 7721..7795 7709..7786 7705..7779 7695..7769 ATP8 7785..7952 7781..7948 7784..7951 7785..7952 7783..7950 7797..7964 7788..7955 7781..7948 7771..7938 ATP6 7943..8626 7939..8622 7942..8625 7943..8626 7941..8624 7955..8638 7946..8629 7939..8622 7929..8612 COX3 8626..9409 8622..9405 8625..9408 8626..9409 8624..9407 8638..9421 8629..9412 8622..9405 8612..9395 tRNAGLY 9410..9478 9406..9474 9409..9477 9410..9478 9408..9476 9422..9490 9413..9481 9406..9474 9396..9464 ND3 9479..9823 9475..9819 9478..9822 9479..9823 9477..9821 9491..9835 9482..9826 9475..9819 9465..9809 tRNAARG 9824..9888 9820..9884 9823..9887 9824..9888 9822..9886 9836..9900 9827..9891 9820..9884 9810..9874 ND4L 9891..10187 9887..10183 9890..10186 9891..10187 9889..10185 9903..10199 9894..10190 9887..10183 9877..10173 ND4 10181..11549 10177..11545 10180..11548 10181..11549 10179..11547 10193..11561 10184..11552 10177..11545 10167..11535 tRNAHIS 11550..11618 11546..11614 11549..11617 11550..11618 11548..11616 11562..11630 11553..11621 11546..11614 11536..11604 tRNASER2 11619..11686 11615..11682 11618..11685 11619..11686 11617..11684 11631..11698 11622..11689 11615..11682 11605..11672 tRNALEU2 11687..11756 11683..11752 11686..11755 11687..11756 11685..11754 11699..11768 11690..11759 11683..11752 11673..11742 ND5 11759..13573 11755..13569 11758..13572 11759..13573 11757..13571 11771..13585 11762..13573 11755..13569 11745..13556 ND6 13569..14084* 13565..14080* 13568..14083* 13569..14084* 13567..14082* 13581..14090* 13569..14084 13565..14080 13552..14067 tRNAGLU 14085..14158* 14081..14154* 14084..14157* 14085..14158* 14083..14156* 14091..14167* 14085..14161* 14081..14157* 14068..14144* CYTB 14163..15303 14159..15299 14162..15302 14163..15303 14161..15301 14172..15300 14166..15306 14162..15302 14149..15307 tRNATHR 15304..15372 15300..15368 15303..15371 15304..15372 15302..15370 15301..15369 15307..15375 15303..15371 15308..15376 NC 15373..16211 15369..16200 15372..16217 15373..16207 15371..16204 15372..15621 15378..15627 15374..15820 15378..15628 tRNAPRO 16212..16280* 16201..16269* 16218..16286* 16208..16276* 16205..16273* 15622..15690* 15628..15696* 15821..15889* 15629..15697* D-loop 16281..17142 16270..17131 16287..17144 16277..17142 16274..17135 15691..16537 15697..16537 15890..16739 15698..16543

Table 3. Nucleotide composition of complete Siren mtDNA sequences from this study. Siren individual Species %A %T %C %G (from this study) S24 Siren spp. (TX) 34.1 32.5 20.2 13.2

SS27 Siren spp. (TX) 34.1 32.4 20.3 13.1

SS6 Siren spp. (TX) 34.2 32.4 20.3 13.1

SS29 Siren spp. (TX) 34.1 32.4 20.3 13.1

SS3 Siren spp. (TX) 34.2 32.5 20.3 13.1

SS20 Siren spp. (TX) 34.4 32.0 20.5 13.1

SS23 Siren spp. (TX) 34.2 31.9 20.6 13.2

ATT-1 Siren spp. (TX) 34.3 32.3 20.3 13.1

SL4 S. lacertina (FL) 34.1 31.7 20.9 13.3

H 64

10 Siren Siren Abundance

0 Sep. Oct. Nov. Dec. May Jun. Jul. Oct. Nov. Mar. 2013 2013 2013 2013 2014 2014 2014 2014 2014 2015 Month Figure 5. Histogram of the total siren abundance collected (regardless of site location) during each month of collection in this study from 2013 to 2015. Abundances display siren collection from systematic sampling, opportunistic captures, and the CPUE study. The abundance did not include one siren (S31) that was retrieved dead.

0.4

0.35

0.3

0.25

0.2

0.15

0.1 CPUE CPUE (siren abundance/trap night) 0.05

0 Sep. Oct. Nov. Dec. May Jun. Jul. Oct. Mar. 2013 2013 2013 2013 2014 2014 2014 2014 2015 Month

Figure 6. Histogram of the catch-per-unit-effort (CPUE) of sirens collected during each month of sampling (regardless of site) in this study from 2013 to 2015. CPUE values display only sirens collected from systematic trapping (not opportunistic captures).

Figure 7. Variation in body coloration and spot pattern of three sirens collected in South Texas in this study.

600

500

400

300 y = 96.934ln(x) - 96.868 2

Length(mm) R = 0.93 200

100

0 0 100 200 300 400 500 Weight (g)

Figure 8. Plot of the weight-length relationships and log-regression for all sirens collected in South Texas from this study. A logarithmic trend- line produced a R2 = 0.93 correlation.

A B

C D

E F F

Figure 9. Siren tail shapes of known Siren lacertina (A, B) and tails shapes of sirens collected in South Texas in this study (C –F).

Figure 10. A small (18.5 g, 185 mm) siren collected from South Texas in October 2014 from this study, exhibiting a distinct yellow pattern along the side of the body.

A B

C D

Figure 11. Variation in siren head spot pattern and coloration of known Siren lacertina (A) and sirens collected in South Texas in this study (B-D).

Figure 12. Wild-ID photograph recognition software analysis for the comparison of the spot patterns on the dorsal region of the head between all sirens within the database. The red box displays the comparison value of 0.0245, which is not considered a match (< 0.100).

Figure 13. Non-metric multidimensional scaling (MDS) ordination plot of the combined percent cover (from the Aquatic, Emergent and Edge zones) for thirty-six sites from May to October 2014 (no samples in August or September) based on the presence/absence of sirens. No significance was observed between siren presence and percent cover. Low stress of 0.08 indicates that the 2D representation of the MDS plot was appropriate. The closer two points are to one another, the more similar the percent cover of all zones is between the sites.

Figure 14. Non-metric multidimensional scaling (MDS) ordination plot of the percent cover in the Edge zone for thirty-six sites from May to October 2014 (no samples in August or September) based on the presence/absence of sirens Low stress of 0.01 indicates that the 2D representation of the MDS plot was appropriate. The closer two points are to one another, the more similar the Edge percent cover is between the sites. The percent edge cover was not significant, but almost all sites with siren collection had ≥ 95% cover.

100 1.0 100 1.0 S. intermedia 100 * 1.0 100 P. axanthus 1.0 100 D. aterrimus 1.0 A. tigrinum

T. natans

100 1.0 97 1.0 87 1.0

1.0

Figure 16. Bayesian and Maximum Likelihood consensus tree of the siren-only complete mitochondrial genome dataset (22tRNAs, 2 rRNAs, 13 proteins, intergenic spacer (IGS), D-Loop). Support values above branches are the Maximum Likelihood bootstrap values (values < 50% not shown) and below the branches are the Bayesian Posterior Probabilities, indicating support for the node.

99 1.0

58 0.99 83 0.99

1.0 S. intermedia P. axanthus

Figure 17. Bayesian and Maximum Likelihood consensus topology tree of the phylogenetic relationships for all Siren spp. from Texas for the 16S dataset. Support values above branches are the Maximum Likelihood bootstrap values (values < 50% not shown) and below the branches are the Bayesian Posterior Probabilities, indicating support for the node.

100 1.0

1.0 50 0.83 S. intermedia P. axanthus

Figure 18. Bayesian and Maximum Likelihood consensus topology tree of the phylogenetic relationships for all Siren spp. from Texas for the CO1 dataset. Support values above branches are the Maximum Likelihood bootstrap values (values < 50% not shown) and below the branches are the Bayesian Posterior Probabilities, indicating support for the node.

99 53 1.0

1.0 1.0

0.99 0.97 0.82

1.0 S. intermedia P. axanthus

Figure 19. Bayesian and Maximum Likelihood consensus topology tree of the phylogenetic relationships for all Siren spp. from Texas for the Cyt-b dataset. Support values above branches are the Maximum Likelihood bootstrap values (values < 50% not shown) and below the branches are the Bayesian Posterior Probabilities, indicating support for the node.

87 0.93

97 1.0

1.0

89 99 0.99 94 1.0 1.0

Figure 20. Bayesian and Maximum Likelihood consensus topology tree of the phylogenetic relationships for all Siren spp. from Texas for the intergenic spacer (IGS) dataset. Support values above branches are the Maximum Likelihood bootstrap values (values < 50% not shown) and below the branches are the Bayesian Posterior Probabilities, indicating support for the node.

y = 1E - 05x + 0.0074 r2 = 0.2801

(km)

Figure 21. The correlation between pairwise comparisons of genetic distance (uncorrected p-values) and geographic distance (km) for the CO1 dataset for all Texas sirens (n = 45) collected in this study. The Mantel Test showed significant correlation with a p-value of 0.02 (p < 0.05). The linear correlation is r2 = 0.28.

COASTAL PLAIN 85

Figure 22. The collection locations in Texas for the eight Texas sirens sequenced for the complete mitochondrial genome. Yellow stars represent the common South Texas genotype, and the colored stars represent the three divergent sirens. The blue star represents the collection location of siren ATT-1, the green star siren SS23, and the red star siren SS20. Due to proximity of location, the eighth star (yellow) is not visible, but is located with the green star. Supplementary Tables and Figures

Table S1. Nine vegetation group categories analyzed for presence/absence within all three zones (edge, emergent, aquatic) in a site. Vegetation groups were analyzed at thirty-six sample sites during the summer of 2014. 1. Cane, reeds, rushes, and cattails 2. Grasses and sedges (< 1 m) 3. Grasses and sedges (> 1 m) 4. Thorny brush (retama and mimosa) 5. Overhanging trees 6. Dead trees, dead logs, parasitic vegetation 7. Aquatic leafy plants 8. Algae 9. Woody trees (willow and mesquite)

Aquatic Emergent Zone Aquatic Submerged Zone Edge Zone

Figure S1. Representation of the three Zone regions that were assessed for percent cover and the presence/absence of vegetation groups at all sampled sites in 2014. The Zone regions include the Edge Zone, the Aquatic Emergent Zone, and the Aquatic Submerged Zone.

Table S2. Common co-occurring species collected in traps in water bodies inhabited by sirens between the sampling season in 2013 and 2014 from this study. Phyla Class Species Common name

Vertebrata Amphibia Rana berlandieri Rio Grande Leopard Frog Reptilia Nerodia rhombifer rhombifer Diamondback water snake

Reptilia Herichthys cyanoguttatus Rio Grande Cichlid Osteichthyes Poecilia formosa Amazon Molly

Osteichthyes Poecilia latipinna Sailfin Molly Osteichthyes Lepomis macrochirus Bluegill sunfish Osteichthyes Lepomis cyanellus Green sunfish

Osteichthyes Cyprinodon veriagatus Sheepshead minnow Osteichthyes Lepomis gulosus Warmouth

Osteichthyes Cyprinus carpio Common Carp

Osteichthyes Hypostomus plecostomus Sucker fish Osteichthyes Oreochromis aureus Blue tilapia Osteichthyes Menidia menidia Silverside Arthropoda Insecta Dytiscidae sp. Predaceous diving beetle

Insecta Belostomatidae sp. Giant water bug Insecta Hydrophilus triangularis Giant Black water beetle

Insecta Ranatra sp. Water stick insects

Insecta Unknown sp. Dragonfly

Malacostraca Procambarus clarkia Red swamp crayfish

Malacostraca Palaemonetes paludosus Glass shrimp Arachnida Dolomedes sp. Raft spider Mollusca Gastropoda Planorbidae sp. Ram’s horn snail

Figure S2. Apparatus and technique used to handle collected sirens in the field for accurate total length measurements. Photograph taken by Seth Patterson.

Table S3. Akaike information criterion (AIC) nucleotide substitution models for the Siren+outgroup and Siren-only gene partitions designated by PartitionFinder. The models of evolution include gamma distributed rate variation among sites (G) and the proportion of invariable sites (I). Parenthesis with numbers indicate the codon position for that applied model.

Siren+outgroup Siren-only tRNAs GTR+G HKY+I 12S GTR+I+G GTR+I 16S GTR+G GTR+G ATP8 GTR+I+G (1,2,3) HKY+I+G (1,2); GTR+G (3) ATP6 GTR+I+G (1,2,3) HKY+I+G (1); HKY+I (2); GTR+G (3) CO1 SYM+I (1); HYK+G (2); GTR+I+G (3) GTR+I (1); HKY+I (2); GTR+G (3) CO2 SYM+I (1); GTR+I (2); GTR+I+G (3) GTR+I (1); HKY+I (2); GTR+G (3) CO3 SYM+I (1); GTR+I (2); GTR+I+G (3) GTR+I (1); HKY+I (2); GTR+G (3) Cyt-b GTR+I+G (1,3); GTR+I (2) GTR+G (1); HKY+G (2); GTR+I+G (3) ND1 GTR+I+G (1,2,3) GTR+I (1); HKY+I (2); GTR+G (3) ND2 GTR+I+G (1,2,3) GTR+G (1); HKY+I (2); GTR+G (3) ND3 GTR+I+G (1,2,3) GTR+I (1); HKY+I(2); GTR+G (3) ND4 GTR+I+G (1,2,3) HKY+I+G (1); HKY+I (2); GTR+G (3) ND4L GTR+I+G (1,3); GTR+G (2) GTR+I (1); HKY+I (2); GTR+G (3) ND5 GTR+G (1); GTR+I+G (2,3) GTR+G (1); HKY+G (2); GTR+I+G (3) ND6 GTR+G (1,2,3) GTR+I (1); HKY+G (2); GRR (3) IGS GTR+G D-Loop HKY+G * AIC model abbreviations: General Time Reversal (GTR), Symmetrical model (SYM), Hasegawa-Kishino-Yano (HKY), Generalized Ridge Regression (GRR).

250

A Historical Records

200

150

100

50 Siren Abundance Siren Abundance in Texas

B 90

80 70 60 50 40 30 20

10 Siren Abundance Siren Abundance in SouthTexas 0 Jan. Feb. Mar. Apr. May Jun. Jul. Aug. Sept. Oct. Nov. Dec. Month

Figure S3. Historical records of siren collections by month (A) throughout the entire state of Texas (B) throughout South Texas (Starr, Hidalgo, Cameron, Willacy, Kleberg, and Jim Wells County). Historical records are documented from 1950 – present and were retrieved from the Texas Natural History Collection (TNHC) and Texas A&M University (TAMU). Siren captures from this study are not included.